1 /* 2 * Copyright (C) 2012 The Android Open Source Project 3 * 4 * Licensed under the Apache License, Version 2.0 (the "License"); 5 * you may not use this file except in compliance with the License. 6 * You may obtain a copy of the License at 7 * 8 * http://www.apache.org/licenses/LICENSE-2.0 9 * 10 * Unless required by applicable law or agreed to in writing, software 11 * distributed under the License is distributed on an "AS IS" BASIS, 12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 13 * See the License for the specific language governing permissions and 14 * limitations under the License. 15 */ 16 17 package android.media; 18 19 import static android.media.codec.Flags.FLAG_IN_PROCESS_SW_AUDIO_CODEC; 20 import static android.media.codec.Flags.FLAG_REGION_OF_INTEREST; 21 22 import static com.android.media.codec.flags.Flags.FLAG_CODEC_IMPORTANCE; 23 import static com.android.media.codec.flags.Flags.FLAG_LARGE_AUDIO_FRAME; 24 25 import android.annotation.FlaggedApi; 26 import android.annotation.IntDef; 27 import android.annotation.NonNull; 28 import android.annotation.Nullable; 29 import android.compat.annotation.UnsupportedAppUsage; 30 import android.graphics.Rect; 31 import android.text.TextUtils; 32 33 import java.lang.annotation.Retention; 34 import java.lang.annotation.RetentionPolicy; 35 import java.nio.ByteBuffer; 36 import java.nio.ByteOrder; 37 import java.util.AbstractSet; 38 import java.util.HashMap; 39 import java.util.Iterator; 40 import java.util.List; 41 import java.util.Map; 42 import java.util.Set; 43 import java.util.stream.Collectors; 44 45 /** 46 * Encapsulates the information describing the format of media data, be it audio or video, as 47 * well as optional feature metadata. 48 * <p> 49 * The format of the media data is specified as key/value pairs. Keys are strings. Values can 50 * be integer, long, float, String or ByteBuffer. 51 * <p> 52 * The feature metadata is specificed as string/boolean pairs. 53 * <p> 54 * Keys common to all audio/video formats, <b>all keys not marked optional are mandatory</b>: 55 * 56 * <table> 57 * <tr><th>Name</th><th>Value Type</th><th>Description</th></tr> 58 * <tr><td>{@link #KEY_MIME}</td><td>String</td><td>The type of the format.</td></tr> 59 * <tr><td>{@link #KEY_CODECS_STRING}</td><td>String</td><td>optional, the RFC 6381 codecs string of the MediaFormat</td></tr> 60 * <tr><td>{@link #KEY_MAX_INPUT_SIZE}</td><td>Integer</td><td>optional, maximum size of a buffer of input data</td></tr> 61 * <tr><td>{@link #KEY_PIXEL_ASPECT_RATIO_WIDTH}</td><td>Integer</td><td>optional, the pixel aspect ratio width</td></tr> 62 * <tr><td>{@link #KEY_PIXEL_ASPECT_RATIO_HEIGHT}</td><td>Integer</td><td>optional, the pixel aspect ratio height</td></tr> 63 * <tr><td>{@link #KEY_BIT_RATE}</td><td>Integer</td><td><b>encoder-only</b>, desired bitrate in bits/second</td></tr> 64 * <tr><td>{@link #KEY_DURATION}</td><td>long</td><td>the duration of the content (in microseconds)</td></tr> 65 * </table> 66 * 67 * Video formats have the following keys: 68 * <table> 69 * <tr><th>Name</th><th>Value Type</th><th>Description</th></tr> 70 * <tr><td>{@link #KEY_WIDTH}</td><td>Integer</td><td></td></tr> 71 * <tr><td>{@link #KEY_HEIGHT}</td><td>Integer</td><td></td></tr> 72 * <tr><td>{@link #KEY_COLOR_FORMAT}</td><td>Integer</td><td>set by the user 73 * for encoders, readable in the output format of decoders</b></td></tr> 74 * <tr><td>{@link #KEY_FRAME_RATE}</td><td>Integer or Float</td><td>required for <b>encoders</b>, 75 * optional for <b>decoders</b></td></tr> 76 * <tr><td>{@link #KEY_CAPTURE_RATE}</td><td>Integer</td><td></td></tr> 77 * <tr><td>{@link #KEY_I_FRAME_INTERVAL}</td><td>Integer (or Float)</td><td><b>encoder-only</b>, 78 * time-interval between key frames. 79 * Float support added in {@link android.os.Build.VERSION_CODES#N_MR1}</td></tr> 80 * <tr><td>{@link #KEY_INTRA_REFRESH_PERIOD}</td><td>Integer</td><td><b>encoder-only</b>, optional</td></tr> 81 * <tr><td>{@link #KEY_LATENCY}</td><td>Integer</td><td><b>encoder-only</b>, optional</td></tr> 82 * <tr><td>{@link #KEY_MAX_WIDTH}</td><td>Integer</td><td><b>decoder-only</b>, optional, max-resolution width</td></tr> 83 * <tr><td>{@link #KEY_MAX_HEIGHT}</td><td>Integer</td><td><b>decoder-only</b>, optional, max-resolution height</td></tr> 84 * <tr><td>{@link #KEY_REPEAT_PREVIOUS_FRAME_AFTER}</td><td>Long</td><td><b>encoder in surface-mode 85 * only</b>, optional</td></tr> 86 * <tr><td>{@link #KEY_PUSH_BLANK_BUFFERS_ON_STOP}</td><td>Integer(1)</td><td><b>decoder rendering 87 * to a surface only</b>, optional</td></tr> 88 * <tr><td>{@link #KEY_TEMPORAL_LAYERING}</td><td>String</td><td><b>encoder only</b>, optional, 89 * temporal-layering schema</td></tr> 90 * </table> 91 * Specify both {@link #KEY_MAX_WIDTH} and {@link #KEY_MAX_HEIGHT} to enable 92 * adaptive playback (seamless resolution change) for a video decoder that 93 * supports it ({@link MediaCodecInfo.CodecCapabilities#FEATURE_AdaptivePlayback}). 94 * The values are used as hints for the codec: they are the maximum expected 95 * resolution to prepare for. Depending on codec support, preparing for larger 96 * maximum resolution may require more memory even if that resolution is never 97 * reached. These fields have no effect for codecs that do not support adaptive 98 * playback.<br /><br /> 99 * 100 * Audio formats have the following keys: 101 * <table> 102 * <tr><th>Name</th><th>Value Type</th><th>Description</th></tr> 103 * <tr><td>{@link #KEY_CHANNEL_COUNT}</td><td>Integer</td><td></td></tr> 104 * <tr><td>{@link #KEY_SAMPLE_RATE}</td><td>Integer</td><td></td></tr> 105 * <tr><td>{@link #KEY_PCM_ENCODING}</td><td>Integer</td><td>optional</td></tr> 106 * <tr><td>{@link #KEY_IS_ADTS}</td><td>Integer</td><td>optional, if <em>decoding</em> AAC audio content, setting this key to 1 indicates that each audio frame is prefixed by the ADTS header.</td></tr> 107 * <tr><td>{@link #KEY_AAC_PROFILE}</td><td>Integer</td><td><b>encoder-only</b>, optional, if content is AAC audio, specifies the desired profile.</td></tr> 108 * <tr><td>{@link #KEY_AAC_SBR_MODE}</td><td>Integer</td><td><b>encoder-only</b>, optional, if content is AAC audio, specifies the desired SBR mode.</td></tr> 109 * <tr><td>{@link #KEY_AAC_DRC_TARGET_REFERENCE_LEVEL}</td><td>Integer</td><td><b>decoder-only</b>, optional, if content is AAC audio, specifies the target reference level.</td></tr> 110 * <tr><td>{@link #KEY_AAC_ENCODED_TARGET_LEVEL}</td><td>Integer</td><td><b>decoder-only</b>, optional, if content is AAC audio, specifies the target reference level used at encoder.</td></tr> 111 * <tr><td>{@link #KEY_AAC_DRC_BOOST_FACTOR}</td><td>Integer</td><td><b>decoder-only</b>, optional, if content is AAC audio, specifies the DRC boost factor.</td></tr> 112 * <tr><td>{@link #KEY_AAC_DRC_ATTENUATION_FACTOR}</td><td>Integer</td><td><b>decoder-only</b>, optional, if content is AAC audio, specifies the DRC attenuation factor.</td></tr> 113 * <tr><td>{@link #KEY_AAC_DRC_HEAVY_COMPRESSION}</td><td>Integer</td><td><b>decoder-only</b>, optional, if content is AAC audio, specifies whether to use heavy compression.</td></tr> 114 * <tr><td>{@link #KEY_AAC_MAX_OUTPUT_CHANNEL_COUNT}</td><td>Integer</td><td><b>decoder-only</b>, optional, if content is AAC audio, specifies the maximum number of channels the decoder outputs.</td></tr> 115 * <tr><td>{@link #KEY_AAC_DRC_EFFECT_TYPE}</td><td>Integer</td><td><b>decoder-only</b>, optional, if content is AAC audio, specifies the MPEG-D DRC effect type to use.</td></tr> 116 * <tr><td>{@link #KEY_AAC_DRC_OUTPUT_LOUDNESS}</td><td>Integer</td><td><b>decoder-only</b>, optional, if content is AAC audio, returns the DRC output loudness.</td></tr> 117 * <tr><td>{@link #KEY_AAC_DRC_ALBUM_MODE}</td><td>Integer</td><td><b>decoder-only</b>, optional, if content is AAC audio, specifies the whether MPEG-D DRC Album Mode is active or not.</td></tr> 118 * <tr><td>{@link #KEY_CHANNEL_MASK}</td><td>Integer</td><td>optional, a mask of audio channel assignments</td></tr> 119 * <tr><td>{@link #KEY_ENCODER_DELAY}</td><td>Integer</td><td>optional, the number of frames to trim from the start of the decoded audio stream.</td></tr> 120 * <tr><td>{@link #KEY_ENCODER_PADDING}</td><td>Integer</td><td>optional, the number of frames to trim from the end of the decoded audio stream.</td></tr> 121 * <tr><td>{@link #KEY_FLAC_COMPRESSION_LEVEL}</td><td>Integer</td><td><b>encoder-only</b>, optional, if content is FLAC audio, specifies the desired compression level.</td></tr> 122 * <tr><td>{@link #KEY_MPEGH_PROFILE_LEVEL_INDICATION}</td><td>Integer</td> 123 * <td><b>decoder-only</b>, optional, if content is MPEG-H audio, 124 * specifies the profile and level of the stream.</td></tr> 125 * <tr><td>{@link #KEY_MPEGH_COMPATIBLE_SETS}</td><td>ByteBuffer</td> 126 * <td><b>decoder-only</b>, optional, if content is MPEG-H audio, 127 * specifies the compatible sets (profile and level) of the stream.</td></tr> 128 * <tr><td>{@link #KEY_MPEGH_REFERENCE_CHANNEL_LAYOUT}</td> 129 * <td>Integer</td><td><b>decoder-only</b>, optional, if content is MPEG-H audio, 130 * specifies the preferred reference channel layout of the stream.</td></tr> 131 * <tr><td>{@link #KEY_MAX_BUFFER_BATCH_OUTPUT_SIZE}</td><td>Integer</td><td>optional, used with 132 * large audio frame support, specifies max size of output buffer in bytes.</td></tr> 133 * <tr><td>{@link #KEY_BUFFER_BATCH_THRESHOLD_OUTPUT_SIZE}</td><td>Integer</td><td>optional, 134 * used with large audio frame support, specifies threshold output size in bytes.</td></tr> 135 * </table> 136 * 137 * Subtitle formats have the following keys: 138 * <table> 139 * <tr><td>{@link #KEY_MIME}</td><td>String</td><td>The type of the format.</td></tr> 140 * <tr><td>{@link #KEY_LANGUAGE}</td><td>String</td><td>The language of the content.</td></tr> 141 * <tr><td>{@link #KEY_CAPTION_SERVICE_NUMBER}</td><td>int</td><td>optional, the closed-caption service or channel number.</td></tr> 142 * </table> 143 * 144 * Image formats have the following keys: 145 * <table> 146 * <tr><td>{@link #KEY_MIME}</td><td>String</td><td>The type of the format.</td></tr> 147 * <tr><td>{@link #KEY_WIDTH}</td><td>Integer</td><td></td></tr> 148 * <tr><td>{@link #KEY_HEIGHT}</td><td>Integer</td><td></td></tr> 149 * <tr><td>{@link #KEY_COLOR_FORMAT}</td><td>Integer</td><td>set by the user 150 * for encoders, readable in the output format of decoders</b></td></tr> 151 * <tr><td>{@link #KEY_TILE_WIDTH}</td><td>Integer</td><td>required if the image has grid</td></tr> 152 * <tr><td>{@link #KEY_TILE_HEIGHT}</td><td>Integer</td><td>required if the image has grid</td></tr> 153 * <tr><td>{@link #KEY_GRID_ROWS}</td><td>Integer</td><td>required if the image has grid</td></tr> 154 * <tr><td>{@link #KEY_GRID_COLUMNS}</td><td>Integer</td><td>required if the image has grid</td></tr> 155 * </table> 156 */ 157 public final class MediaFormat { 158 public static final String MIMETYPE_VIDEO_VP8 = "video/x-vnd.on2.vp8"; 159 public static final String MIMETYPE_VIDEO_VP9 = "video/x-vnd.on2.vp9"; 160 public static final String MIMETYPE_VIDEO_AV1 = "video/av01"; 161 public static final String MIMETYPE_VIDEO_AVC = "video/avc"; 162 public static final String MIMETYPE_VIDEO_HEVC = "video/hevc"; 163 public static final String MIMETYPE_VIDEO_MPEG4 = "video/mp4v-es"; 164 public static final String MIMETYPE_VIDEO_H263 = "video/3gpp"; 165 public static final String MIMETYPE_VIDEO_MPEG2 = "video/mpeg2"; 166 public static final String MIMETYPE_VIDEO_RAW = "video/raw"; 167 public static final String MIMETYPE_VIDEO_DOLBY_VISION = "video/dolby-vision"; 168 public static final String MIMETYPE_VIDEO_SCRAMBLED = "video/scrambled"; 169 170 public static final String MIMETYPE_AUDIO_AMR_NB = "audio/3gpp"; 171 public static final String MIMETYPE_AUDIO_AMR_WB = "audio/amr-wb"; 172 public static final String MIMETYPE_AUDIO_MPEG = "audio/mpeg"; 173 public static final String MIMETYPE_AUDIO_AAC = "audio/mp4a-latm"; 174 public static final String MIMETYPE_AUDIO_QCELP = "audio/qcelp"; 175 public static final String MIMETYPE_AUDIO_VORBIS = "audio/vorbis"; 176 public static final String MIMETYPE_AUDIO_OPUS = "audio/opus"; 177 public static final String MIMETYPE_AUDIO_G711_ALAW = "audio/g711-alaw"; 178 public static final String MIMETYPE_AUDIO_G711_MLAW = "audio/g711-mlaw"; 179 public static final String MIMETYPE_AUDIO_RAW = "audio/raw"; 180 public static final String MIMETYPE_AUDIO_FLAC = "audio/flac"; 181 public static final String MIMETYPE_AUDIO_MSGSM = "audio/gsm"; 182 public static final String MIMETYPE_AUDIO_AC3 = "audio/ac3"; 183 public static final String MIMETYPE_AUDIO_EAC3 = "audio/eac3"; 184 public static final String MIMETYPE_AUDIO_EAC3_JOC = "audio/eac3-joc"; 185 public static final String MIMETYPE_AUDIO_AC4 = "audio/ac4"; 186 public static final String MIMETYPE_AUDIO_SCRAMBLED = "audio/scrambled"; 187 /** MIME type for MPEG-H Audio single stream */ 188 public static final String MIMETYPE_AUDIO_MPEGH_MHA1 = "audio/mha1"; 189 /** MIME type for MPEG-H Audio single stream, encapsulated in MHAS */ 190 public static final String MIMETYPE_AUDIO_MPEGH_MHM1 = "audio/mhm1"; 191 /** MIME type for DTS Digital Surround (up to 5.1 channels) audio stream, aka DTS-CA. */ 192 public static final String MIMETYPE_AUDIO_DTS = "audio/vnd.dts"; 193 /** 194 * MIME type for DTS HD (up to 7.1 channels) audio stream. 195 * With codec profile DTS_HDProfileHRA represents DTS HD High Resolution Audio. 196 * With codec profile DTS_HDProfileMA represents DTS HD Master Audio. 197 * With codec profile DTS_HDProfileLBR represents DTS Express. 198 */ 199 public static final String MIMETYPE_AUDIO_DTS_HD = "audio/vnd.dts.hd"; 200 /** 201 * MIME type for DTS UHD (object-based) audio stream, aka DTS:X. 202 * With codec profile DTS_UHDProfileP1 represents DTS-UHD P1. 203 * With codec profile DTS_UHDProfileP2 represents DTS-UHD P2. 204 */ 205 public static final String MIMETYPE_AUDIO_DTS_UHD = "audio/vnd.dts.uhd"; 206 /** MIME type for Dynamic Resolution Adaptation (DRA) audio stream. */ 207 public static final String MIMETYPE_AUDIO_DRA = "audio/vnd.dra"; 208 /** MIME type for Dolby Metadata-enhanced Audio Transmission (MAT) audio stream. */ 209 public static final String MIMETYPE_AUDIO_DOLBY_MAT = "audio/vnd.dolby.mat"; 210 /** MIME type for Dolby TrueHD audio format, based on Meridian Lossless Packing (MLP). */ 211 public static final String MIMETYPE_AUDIO_DOLBY_TRUEHD = "audio/vnd.dolby.mlp"; 212 /** 213 * MIME type for AAC Low Complexity (LC) audio stream. Uses the scheme defined by 214 * RFC 6381 with OTI of MPEG-4 (40) and AOT of AAC LC (2) from ISO/IEC 14496-3. 215 */ 216 public static final String MIMETYPE_AUDIO_AAC_LC = "audio/mp4a.40.02"; 217 /** 218 * MIME type for HE-AAC v1 (LC + SBR) audio stream. Uses the scheme defined by 219 * RFC 6381 with OTI of MPEG-4 (40) and AOT of AAC SBR (5) from ISO/IEC 14496-3. 220 */ 221 public static final String MIMETYPE_AUDIO_AAC_HE_V1 = "audio/mp4a.40.05"; 222 /** 223 * MIME type for HE-AAC v2 (LC + SBR + PS) audio stream. Uses the scheme defined by 224 * RFC 6381 with OTI of MPEG-4 (40) and AOT of PS (29) from ISO/IEC 14496-3. 225 */ 226 public static final String MIMETYPE_AUDIO_AAC_HE_V2 = "audio/mp4a.40.29"; 227 /** 228 * MIME type for AAC Enhanced Low Delay (ELD) audio stream. Uses the scheme defined by 229 * RFC 6381 with OTI of MPEG-4 (40) and AOT of ELD (39) from ISO/IEC 14496-3. 230 */ 231 public static final String MIMETYPE_AUDIO_AAC_ELD = "audio/mp4a.40.39"; 232 /** 233 * MIME type for AAC XHE audio stream. Uses the scheme defined by 234 * RFC 6381 with OTI of MPEG-4 (40) and AOT of USAC (42) from ISO/IEC 14496-3. 235 */ 236 public static final String MIMETYPE_AUDIO_AAC_XHE = "audio/mp4a.40.42"; 237 /** 238 * MIME type for MPEG-H Baseline (BL) Profile L3 audio stream. Uses the scheme defined by 239 * RFC 6381 with mpegh3daProfileLevelIndication for main profile/L3 (0x3) from ISO/IEC 23008-3. 240 */ 241 public static final String MIMETYPE_AUDIO_MPEGH_BL_L3 = "audio/mhm1.03"; 242 /** 243 * MIME type for MPEG-H Baseline (BL) Profile L4 audio stream. Uses the scheme defined by 244 * RFC 6381 with mpegh3daProfileLevelIndication for main profile/L4 (0x4) from ISO/IEC 23008-3. 245 */ 246 public static final String MIMETYPE_AUDIO_MPEGH_BL_L4 = "audio/mhm1.04"; 247 /** 248 * MIME type for MPEG-H Low Complexity (LC) L3 audio stream. Uses the scheme defined by 249 * RFC 6381 with mpegh3daProfileLevelIndication for LC profile/L3 (0xD) from ISO/IEC 23008-3. 250 */ 251 public static final String MIMETYPE_AUDIO_MPEGH_LC_L3 = "audio/mhm1.0d"; 252 /** 253 * MIME type for MPEG-H Low Complexity (LC) L4 audio stream. Uses the scheme defined by 254 * RFC 6381 with mpegh3daProfileLevelIndication for LC profile/L4 (0xE) from ISO/IEC 23008-3. 255 */ 256 public static final String MIMETYPE_AUDIO_MPEGH_LC_L4 = "audio/mhm1.0e"; 257 /** 258 * MIME type for the IEC61937 audio stream encapsulation. This type isn't defined by IANA. 259 */ 260 public static final String MIMETYPE_AUDIO_IEC61937 = "audio/x-iec61937"; 261 262 /** 263 * MIME type for HEIF still image data encoded in HEVC. 264 * 265 * To decode such an image, {@link MediaCodec} decoder for 266 * {@link #MIMETYPE_VIDEO_HEVC} shall be used. The client needs to form 267 * the correct {@link #MediaFormat} based on additional information in 268 * the track format (shown in the next paragraph), and send it to 269 * {@link MediaCodec#configure}. 270 * 271 * The track's MediaFormat will come with {@link #KEY_WIDTH} and 272 * {@link #KEY_HEIGHT} keys, which describes the width and height 273 * of the image. If the image doesn't contain grid (i.e. none of 274 * {@link #KEY_TILE_WIDTH}, {@link #KEY_TILE_HEIGHT}, 275 * {@link #KEY_GRID_ROWS}, {@link #KEY_GRID_COLUMNS} are present), the 276 * track will contain a single sample of coded data for the entire image, 277 * and the image width and height should be used to set up the decoder. 278 * 279 * If the image does come with grid, each sample from the track will 280 * contain one tile in the grid, of which the size is described by 281 * {@link #KEY_TILE_WIDTH} and {@link #KEY_TILE_HEIGHT}. This size 282 * (instead of {@link #KEY_WIDTH} and {@link #KEY_HEIGHT}) should be 283 * used to set up the decoder. The track contains {@link #KEY_GRID_ROWS} 284 * by {@link #KEY_GRID_COLUMNS} samples in row-major, top-row first, 285 * left-to-right order. The output image should be reconstructed by 286 * first tiling the decoding results of the tiles in the correct order, 287 * then trimming (before rotation is applied) on the bottom and right 288 * side, if the tiled area is larger than the image width and height. 289 */ 290 public static final String MIMETYPE_IMAGE_ANDROID_HEIC = "image/vnd.android.heic"; 291 292 /** 293 * MIME type for AVIF still image data encoded in AV1. 294 * 295 * To decode such an image, {@link MediaCodec} decoder for 296 * {@link #MIMETYPE_VIDEO_AV1} shall be used. The client needs to form 297 * the correct {@link #MediaFormat} based on additional information in 298 * the track format (shown in the next paragraph), and send it to 299 * {@link MediaCodec#configure}. 300 * 301 * The track's MediaFormat will come with {@link #KEY_WIDTH} and 302 * {@link #KEY_HEIGHT} keys, which describes the width and height 303 * of the image. If the image doesn't contain grid (i.e. none of 304 * {@link #KEY_TILE_WIDTH}, {@link #KEY_TILE_HEIGHT}, 305 * {@link #KEY_GRID_ROWS}, {@link #KEY_GRID_COLUMNS} are present), the 306 * track will contain a single sample of coded data for the entire image, 307 * and the image width and height should be used to set up the decoder. 308 * 309 * If the image does come with grid, each sample from the track will 310 * contain one tile in the grid, of which the size is described by 311 * {@link #KEY_TILE_WIDTH} and {@link #KEY_TILE_HEIGHT}. This size 312 * (instead of {@link #KEY_WIDTH} and {@link #KEY_HEIGHT}) should be 313 * used to set up the decoder. The track contains {@link #KEY_GRID_ROWS} 314 * by {@link #KEY_GRID_COLUMNS} samples in row-major, top-row first, 315 * left-to-right order. The output image should be reconstructed by 316 * first tiling the decoding results of the tiles in the correct order, 317 * then trimming (before rotation is applied) on the bottom and right 318 * side, if the tiled area is larger than the image width and height. 319 */ 320 public static final String MIMETYPE_IMAGE_AVIF = "image/avif"; 321 322 /** 323 * MIME type for WebVTT subtitle data. 324 */ 325 public static final String MIMETYPE_TEXT_VTT = "text/vtt"; 326 327 /** 328 * MIME type for SubRip (SRT) container. 329 */ 330 public static final String MIMETYPE_TEXT_SUBRIP = "application/x-subrip"; 331 332 /** 333 * MIME type for CEA-608 closed caption data. 334 */ 335 public static final String MIMETYPE_TEXT_CEA_608 = "text/cea-608"; 336 337 /** 338 * MIME type for CEA-708 closed caption data. 339 */ 340 public static final String MIMETYPE_TEXT_CEA_708 = "text/cea-708"; 341 342 @UnsupportedAppUsage 343 private Map<String, Object> mMap; 344 345 /** 346 * A key describing the log session ID for MediaCodec. The log session ID is a random 32-byte 347 * hexadecimal string that is used to associate metrics from multiple media codec instances 348 * to the same playback or recording session. The value is created as 349 * {@link android.media.metrics.LogSessionId LogSessionId}. Sessions are created in 350 * {@link android.media.metrics.MediaMetricsManager MediaMetricsManager}. 351 * The associated value is a string. 352 */ 353 public static final String LOG_SESSION_ID = "log-session-id"; 354 355 /** 356 * A key describing the mime type of the MediaFormat. 357 * The associated value is a string. 358 */ 359 public static final String KEY_MIME = "mime"; 360 361 /** 362 * A key describing the codecs string of the MediaFormat. See RFC 6381 section 3.2 for the 363 * syntax of the value. The value does not hold {@link MediaCodec}-exposed codec names. 364 * The associated value is a string. 365 * 366 * @see MediaParser.TrackData#mediaFormat 367 */ 368 public static final String KEY_CODECS_STRING = "codecs-string"; 369 370 /** 371 * An optional key describing the low latency decoding mode. This is an optional parameter 372 * that applies only to decoders. If enabled, the decoder doesn't hold input and output 373 * data more than required by the codec standards. 374 * The associated value is an integer (0 or 1): 1 when low-latency decoding is enabled, 375 * 0 otherwise. The default value is 0. 376 */ 377 public static final String KEY_LOW_LATENCY = "low-latency"; 378 379 /** 380 * A key describing the language of the content, using either ISO 639-1 381 * or 639-2/T codes. The associated value is a string. 382 */ 383 public static final String KEY_LANGUAGE = "language"; 384 385 /** 386 * A key describing the closed caption service number. For CEA-608 caption tracks, holds the 387 * channel number. For CEA-708, holds the service number. 388 * The associated value is an int. 389 */ 390 public static final String KEY_CAPTION_SERVICE_NUMBER = "caption-service-number"; 391 392 /** 393 * A key describing the sample rate of an audio format. 394 * The associated value is an integer 395 */ 396 public static final String KEY_SAMPLE_RATE = "sample-rate"; 397 398 /** 399 * A key describing the number of channels in an audio format. 400 * The associated value is an integer 401 */ 402 public static final String KEY_CHANNEL_COUNT = "channel-count"; 403 404 /** 405 * A key describing the width of the content in a video format. 406 * The associated value is an integer 407 */ 408 public static final String KEY_WIDTH = "width"; 409 410 /** 411 * A key describing the height of the content in a video format. 412 * The associated value is an integer 413 */ 414 public static final String KEY_HEIGHT = "height"; 415 416 /** 417 * A key describing the bottom-coordinate (y) of the crop rectangle. 418 * This is the bottom-most row included in the crop frame, 419 * where row indices start at 0. 420 * Additional information on the crop rectangle semantics can be found at 421 * {@link android.media.MediaCodec}. 422 */ 423 public static final String KEY_CROP_BOTTOM = "crop-bottom"; 424 425 /** 426 * A key describing the left-coordinate (x) of the crop rectangle. 427 * This is the left-most column included in the crop frame, 428 * where column indices start at 0. 429 * Additional information on the crop rectangle semantics can be found at 430 * {@link android.media.MediaCodec}. 431 */ 432 public static final String KEY_CROP_LEFT = "crop-left"; 433 434 /** 435 * A key describing the right-coordinate (x) of the crop rectangle. 436 * This is the right-most column included in the crop frame, 437 * where column indices start at 0. 438 * Additional information on the crop rectangle semantics can be found at 439 * {@link android.media.MediaCodec}. 440 */ 441 public static final String KEY_CROP_RIGHT = "crop-right"; 442 443 /** 444 * A key describing the top-coordinate (y) of the crop rectangle. 445 * This is the top-most row included in the crop frame, 446 * where row indices start at 0. 447 * Additional information on the crop rectangle semantics can be found at 448 * {@link android.media.MediaCodec}. 449 */ 450 public static final String KEY_CROP_TOP = "crop-top"; 451 452 /** 453 * A key describing the maximum expected width of the content in a video 454 * decoder format, in case there are resolution changes in the video content. 455 * The associated value is an integer 456 */ 457 public static final String KEY_MAX_WIDTH = "max-width"; 458 459 /** 460 * A key describing the maximum expected height of the content in a video 461 * decoder format, in case there are resolution changes in the video content. 462 * The associated value is an integer 463 */ 464 public static final String KEY_MAX_HEIGHT = "max-height"; 465 466 /** A key describing the maximum size in bytes of a buffer of data 467 * described by this MediaFormat. 468 * The associated value is an integer 469 */ 470 public static final String KEY_MAX_INPUT_SIZE = "max-input-size"; 471 472 /** 473 * A key describing the maximum output buffer size in bytes when using 474 * large buffer mode containing multiple access units. 475 * 476 * When not-set - codec functions with one access-unit per frame. 477 * When set less than the size of two access-units - will make codec 478 * operate in single access-unit per output frame. 479 * When set to a value too big - The component or the framework will 480 * override this value to a reasonable max size not exceeding typical 481 * 10 seconds of data (device dependent) when set to a value larger than 482 * that. The value final value used will be returned in the output format. 483 * 484 * The associated value is an integer 485 * 486 * @see FEATURE_MultipleFrames 487 */ 488 @FlaggedApi(FLAG_LARGE_AUDIO_FRAME) 489 public static final String KEY_BUFFER_BATCH_MAX_OUTPUT_SIZE = "buffer-batch-max-output-size"; 490 491 /** 492 * A key describing the threshold output size in bytes when using large buffer 493 * mode containing multiple access units. 494 * 495 * This is an optional parameter. 496 * 497 * If not set - the component can set this to a reasonable value. 498 * If set larger than max size, the components will 499 * clip this setting to maximum buffer batching output size. 500 * 501 * The component will return a partial output buffer if the output buffer reaches or 502 * surpass this limit. 503 * 504 * Threshold size should be always less or equal to KEY_MAX_BUFFER_BATCH_OUTPUT_SIZE. 505 * The final setting of this value as determined by the component will be returned 506 * in the output format 507 * 508 * The associated value is an integer 509 * 510 * @see FEATURE_MultipleFrames 511 */ 512 @FlaggedApi(FLAG_LARGE_AUDIO_FRAME) 513 public static final String KEY_BUFFER_BATCH_THRESHOLD_OUTPUT_SIZE = 514 "buffer-batch-threshold-output-size"; 515 516 /** 517 * A key describing the pixel aspect ratio width. 518 * The associated value is an integer 519 */ 520 public static final String KEY_PIXEL_ASPECT_RATIO_WIDTH = "sar-width"; 521 522 /** 523 * A key describing the pixel aspect ratio height. 524 * The associated value is an integer 525 */ 526 public static final String KEY_PIXEL_ASPECT_RATIO_HEIGHT = "sar-height"; 527 528 /** 529 * A key describing the average bitrate in bits/sec. 530 * The associated value is an integer 531 */ 532 public static final String KEY_BIT_RATE = "bitrate"; 533 534 /** 535 * A key describing the hardware AV sync id. 536 * The associated value is an integer 537 * 538 * See android.media.tv.tuner.Tuner#getAvSyncHwId. 539 */ 540 public static final String KEY_HARDWARE_AV_SYNC_ID = "hw-av-sync-id"; 541 542 /** 543 * A key describing the max bitrate in bits/sec. 544 * This is usually over a one-second sliding window (e.g. over any window of one second). 545 * The associated value is an integer 546 * @hide 547 */ 548 public static final String KEY_MAX_BIT_RATE = "max-bitrate"; 549 550 /** 551 * A key describing the color format of the content in a video format. 552 * Constants are declared in {@link android.media.MediaCodecInfo.CodecCapabilities}. 553 */ 554 public static final String KEY_COLOR_FORMAT = "color-format"; 555 556 /** 557 * A key describing the frame rate of a video format in frames/sec. 558 * <p> 559 * The associated value is normally an integer when the value is used by the platform, 560 * but video codecs also accept float configuration values. 561 * Specifically, {@link MediaExtractor#getTrackFormat MediaExtractor} provides an integer 562 * value corresponding to the frame rate information of the track if specified and non-zero. 563 * Otherwise, this key is not present. {@link MediaCodec#configure MediaCodec} accepts both 564 * float and integer values. 565 * <p> 566 * This represents the desired operating frame rate if the 567 * {@link #KEY_OPERATING_RATE} is not present and {@link #KEY_PRIORITY} is {@code 0} 568 * (realtime). Otherwise, this is just informational. 569 * <p> 570 * For video encoders this value corresponds to the intended frame rate (the rate at which 571 * the application intends to send frames to the encoder, as calculated by the buffer 572 * timestamps, and not from the actual real-time rate that the frames are sent to 573 * the encoder). Encoders use this hint for rate control, specifically for the initial 574 * frames, as encoders are expected to support variable frame rate (for rate control) based 575 * on the actual {@link MediaCodec.BufferInfo#presentationTimeUs buffer timestamps} of 576 * subsequent frames. 577 * <p> 578 * This key is not used in the {@code MediaCodec} 579 * {@link MediaCodec#getInputFormat input}/{@link MediaCodec#getOutputFormat output} formats, 580 * nor by {@link MediaMuxer#addTrack MediaMuxer}. 581 */ 582 public static final String KEY_FRAME_RATE = "frame-rate"; 583 584 /** 585 * A key describing the width (in pixels) of each tile of the content in a 586 * {@link #MIMETYPE_IMAGE_ANDROID_HEIC} / {@link #MIMETYPE_IMAGE_AVIF} track. 587 * The associated value is an integer. 588 * 589 * Refer to {@link #MIMETYPE_IMAGE_ANDROID_HEIC} / {@link #MIMETYPE_IMAGE_AVIF} on decoding 590 * instructions of such tracks. 591 * 592 * @see #KEY_TILE_HEIGHT 593 * @see #KEY_GRID_ROWS 594 * @see #KEY_GRID_COLUMNS 595 */ 596 public static final String KEY_TILE_WIDTH = "tile-width"; 597 598 /** 599 * A key describing the height (in pixels) of each tile of the content in a 600 * {@link #MIMETYPE_IMAGE_ANDROID_HEIC} / {@link #MIMETYPE_IMAGE_AVIF} track. 601 * The associated value is an integer. 602 * 603 * Refer to {@link #MIMETYPE_IMAGE_ANDROID_HEIC} / {@link #MIMETYPE_IMAGE_AVIF} on decoding 604 * instructions of such tracks. 605 * 606 * @see #KEY_TILE_WIDTH 607 * @see #KEY_GRID_ROWS 608 * @see #KEY_GRID_COLUMNS 609 */ 610 public static final String KEY_TILE_HEIGHT = "tile-height"; 611 612 /** 613 * A key describing the number of grid rows in the content in a 614 * {@link #MIMETYPE_IMAGE_ANDROID_HEIC} / {@link #MIMETYPE_IMAGE_AVIF} track. 615 * The associated value is an integer. 616 * 617 * Refer to {@link #MIMETYPE_IMAGE_ANDROID_HEIC} / {@link #MIMETYPE_IMAGE_AVIF} on decoding 618 * instructions of such tracks. 619 * 620 * @see #KEY_TILE_WIDTH 621 * @see #KEY_TILE_HEIGHT 622 * @see #KEY_GRID_COLUMNS 623 */ 624 public static final String KEY_GRID_ROWS = "grid-rows"; 625 626 /** 627 * A key describing the number of grid columns in the content in a 628 * {@link #MIMETYPE_IMAGE_ANDROID_HEIC} / {@link #MIMETYPE_IMAGE_AVIF} track. 629 * The associated value is an integer. 630 * 631 * Refer to {@link #MIMETYPE_IMAGE_ANDROID_HEIC} / {@link #MIMETYPE_IMAGE_AVIF} on decoding 632 * instructions of such tracks. 633 * 634 * @see #KEY_TILE_WIDTH 635 * @see #KEY_TILE_HEIGHT 636 * @see #KEY_GRID_ROWS 637 */ 638 public static final String KEY_GRID_COLUMNS = "grid-cols"; 639 640 /** 641 * A key describing the raw audio sample encoding/format. 642 * 643 * <p>The associated value is an integer, using one of the 644 * {@link AudioFormat}.ENCODING_PCM_ values.</p> 645 * 646 * <p>This is an optional key for audio decoders and encoders specifying the 647 * desired raw audio sample format during {@link MediaCodec#configure 648 * MediaCodec.configure(…)} call. Use {@link MediaCodec#getInputFormat 649 * MediaCodec.getInput}/{@link MediaCodec#getOutputFormat OutputFormat(…)} 650 * to confirm the actual format. For the PCM decoder this key specifies both 651 * input and output sample encodings.</p> 652 * 653 * <p>This key is also used by {@link MediaExtractor} to specify the sample 654 * format of audio data, if it is specified.</p> 655 * 656 * <p>If this key is missing, the raw audio sample format is signed 16-bit short.</p> 657 */ 658 public static final String KEY_PCM_ENCODING = "pcm-encoding"; 659 660 /** 661 * A key describing the capture rate of a video format in frames/sec. 662 * <p> 663 * When capture rate is different than the frame rate, it means that the 664 * video is acquired at a different rate than the playback, which produces 665 * slow motion or timelapse effect during playback. Application can use the 666 * value of this key to tell the relative speed ratio between capture and 667 * playback rates when the video was recorded. 668 * </p> 669 * <p> 670 * The associated value is an integer or a float. 671 * </p> 672 */ 673 public static final String KEY_CAPTURE_RATE = "capture-rate"; 674 675 /** 676 * A key for retrieving the slow-motion marker information associated with a video track. 677 * <p> 678 * The associated value is a ByteBuffer in {@link ByteOrder#BIG_ENDIAN} 679 * (networking order) of the following format: 680 * </p> 681 * <pre class="prettyprint"> 682 * float(32) playbackRate; 683 * unsigned int(32) numMarkers; 684 * for (i = 0;i < numMarkers; i++) { 685 * int(64) timestampUs; 686 * float(32) speedRatio; 687 * }</pre> 688 * The meaning of each field is as follows: 689 * <table border="1" width="90%" align="center" cellpadding="5"> 690 * <tbody> 691 * <tr> 692 * <td>playbackRate</td> 693 * <td>The frame rate at which the playback should happen (or the flattened 694 * clip should be).</td> 695 * </tr> 696 * <tr> 697 * <td>numMarkers</td> 698 * <td>The number of slow-motion markers that follows.</td> 699 * </tr> 700 * <tr> 701 * <td>timestampUs</td> 702 * <td>The starting point of a new segment.</td> 703 * </tr> 704 * <tr> 705 * <td>speedRatio</td> 706 * <td>The playback speed for that segment. The playback speed is a floating 707 * point number, indicating how fast the time progresses relative to that 708 * written in the container. (Eg. 4.0 means time goes 4x as fast, which 709 * makes 30fps become 120fps.)</td> 710 * </tr> 711 * </table> 712 * <p> 713 * The following constraints apply to the timestampUs of the markers: 714 * </p> 715 * <li>The timestampUs shall be monotonically increasing.</li> 716 * <li>The timestampUs shall fall within the time span of the video track.</li> 717 * <li>The first timestampUs should match that of the first video sample.</li> 718 */ 719 public static final String KEY_SLOW_MOTION_MARKERS = "slow-motion-markers"; 720 721 /** 722 * A key describing the frequency of key frames expressed in seconds between key frames. 723 * <p> 724 * This key is used by video encoders. 725 * A negative value means no key frames are requested after the first frame. 726 * A zero value means a stream containing all key frames is requested. 727 * <p class=note> 728 * Most video encoders will convert this value of the number of non-key-frames between 729 * key-frames, using the {@linkplain #KEY_FRAME_RATE frame rate} information; therefore, 730 * if the actual frame rate differs (e.g. input frames are dropped or the frame rate 731 * changes), the <strong>time interval</strong> between key frames will not be the 732 * configured value. 733 * <p> 734 * The associated value is an integer (or float since 735 * {@link android.os.Build.VERSION_CODES#N_MR1}). 736 */ 737 public static final String KEY_I_FRAME_INTERVAL = "i-frame-interval"; 738 739 /** 740 * An optional key describing the period of intra refresh in frames. This is an 741 * optional parameter that applies only to video encoders. If encoder supports it 742 * ({@link MediaCodecInfo.CodecCapabilities#FEATURE_IntraRefresh}), the whole 743 * frame is completely refreshed after the specified period. Also for each frame, 744 * a fix subset of macroblocks must be intra coded which leads to more constant bitrate 745 * than inserting a key frame. This key is recommended for video streaming applications 746 * as it provides low-delay and good error-resilience. This key is ignored if the 747 * video encoder does not support the intra refresh feature. Use the output format to 748 * verify that this feature was enabled. 749 * The associated value is an integer. 750 */ 751 public static final String KEY_INTRA_REFRESH_PERIOD = "intra-refresh-period"; 752 753 /** 754 * An optional key describing whether encoders prepend headers to sync frames (e.g. 755 * SPS and PPS to IDR frames for H.264). This is an optional parameter that applies only 756 * to video encoders. A video encoder may not support this feature; the component will fail 757 * to configure in that case. For other components, this key is ignored. 758 * 759 * The value is an integer, with 1 indicating to prepend headers to every sync frames, 760 * or 0 otherwise. The default value is 0. 761 */ 762 public static final String KEY_PREPEND_HEADER_TO_SYNC_FRAMES = "prepend-sps-pps-to-idr-frames"; 763 764 /** 765 * A key describing the temporal layering schema. This is an optional parameter 766 * that applies only to video encoders. Use {@link MediaCodec#getOutputFormat} 767 * after {@link MediaCodec#configure configure} to query if the encoder supports 768 * the desired schema. Supported values are {@code webrtc.vp8.N-layer}, 769 * {@code android.generic.N}, {@code android.generic.N+M} and {@code none}, where 770 * {@code N} denotes the total number of non-bidirectional layers (which must be at least 1) 771 * and {@code M} denotes the total number of bidirectional layers (which must be non-negative). 772 * <p class=note>{@code android.generic.*} schemas have been added in {@link 773 * android.os.Build.VERSION_CODES#N_MR1}. 774 * <p> 775 * The encoder may support fewer temporal layers, in which case the output format 776 * will contain the configured schema. If the encoder does not support temporal 777 * layering, the output format will not have an entry with this key. 778 * The associated value is a string. 779 */ 780 public static final String KEY_TEMPORAL_LAYERING = "ts-schema"; 781 782 /** 783 * A key describing the stride of the video bytebuffer layout. 784 * Stride (or row increment) is the difference between the index of a pixel 785 * and that of the pixel directly underneath. For YUV 420 formats, the 786 * stride corresponds to the Y plane; the stride of the U and V planes can 787 * be calculated based on the color format, though it is generally undefined 788 * and depends on the device and release. 789 * The associated value is an integer, representing number of bytes. 790 */ 791 public static final String KEY_STRIDE = "stride"; 792 793 /** 794 * A key describing the plane height of a multi-planar (YUV) video bytebuffer layout. 795 * Slice height (or plane height/vertical stride) is the number of rows that must be skipped 796 * to get from the top of the Y plane to the top of the U plane in the bytebuffer. In essence 797 * the offset of the U plane is sliceHeight * stride. The height of the U/V planes 798 * can be calculated based on the color format, though it is generally undefined 799 * and depends on the device and release. 800 * The associated value is an integer, representing number of rows. 801 */ 802 public static final String KEY_SLICE_HEIGHT = "slice-height"; 803 804 /** 805 * Applies only when configuring a video encoder in "surface-input" mode. 806 * The associated value is a long and gives the time in microseconds 807 * after which the frame previously submitted to the encoder will be 808 * repeated (once) if no new frame became available since. 809 */ 810 public static final String KEY_REPEAT_PREVIOUS_FRAME_AFTER 811 = "repeat-previous-frame-after"; 812 813 /** 814 * Instruct the video encoder in "surface-input" mode to drop excessive 815 * frames from the source, so that the input frame rate to the encoder 816 * does not exceed the specified fps. 817 * 818 * The associated value is a float, representing the max frame rate to 819 * feed the encoder at. 820 * 821 */ 822 public static final String KEY_MAX_FPS_TO_ENCODER 823 = "max-fps-to-encoder"; 824 825 /** 826 * Instruct the video encoder in "surface-input" mode to limit the gap of 827 * timestamp between any two adjacent frames fed to the encoder to the 828 * specified amount (in micro-second). 829 * 830 * The associated value is a long int. When positive, it represents the max 831 * timestamp gap between two adjacent frames fed to the encoder. When negative, 832 * the absolute value represents a fixed timestamp gap between any two adjacent 833 * frames fed to the encoder. Note that this will also apply even when the 834 * original timestamp goes backward in time. Under normal conditions, such frames 835 * would be dropped and not sent to the encoder. 836 * 837 * The output timestamp will be restored to the original timestamp and will 838 * not be affected. 839 * 840 * This is used in some special scenarios where input frames arrive sparingly 841 * but it's undesirable to allocate more bits to any single frame, or when it's 842 * important to ensure all frames are captured (rather than captured in the 843 * correct order). 844 * 845 */ 846 public static final String KEY_MAX_PTS_GAP_TO_ENCODER 847 = "max-pts-gap-to-encoder"; 848 849 /** 850 * If specified when configuring a video encoder that's in "surface-input" 851 * mode, it will instruct the encoder to put the surface source in suspended 852 * state when it's connected. No video frames will be accepted until a resume 853 * operation (see {@link MediaCodec#PARAMETER_KEY_SUSPEND}), optionally with 854 * timestamp specified via {@link MediaCodec#PARAMETER_KEY_SUSPEND_TIME}, is 855 * received. 856 * 857 * The value is an integer, with 1 indicating to create with the surface 858 * source suspended, or 0 otherwise. The default value is 0. 859 * 860 * If this key is not set or set to 0, the surface source will accept buffers 861 * as soon as it's connected to the encoder (although they may not be encoded 862 * immediately). This key can be used when the client wants to prepare the 863 * encoder session in advance, but do not want to accept buffers immediately. 864 */ 865 public static final String KEY_CREATE_INPUT_SURFACE_SUSPENDED 866 = "create-input-buffers-suspended"; 867 868 /** 869 * If specified when configuring a video decoder rendering to a surface, 870 * causes the decoder to output "blank", i.e. black frames to the surface 871 * when stopped to clear out any previously displayed contents. 872 * The associated value is an integer of value 1. 873 */ 874 public static final String KEY_PUSH_BLANK_BUFFERS_ON_STOP 875 = "push-blank-buffers-on-shutdown"; 876 877 /** 878 * A key describing the duration (in microseconds) of the content. 879 * The associated value is a long. 880 */ 881 public static final String KEY_DURATION = "durationUs"; 882 883 /** 884 * A key mapping to a value of 1 if the content is AAC audio and 885 * audio frames are prefixed with an ADTS header. 886 * The associated value is an integer (0 or 1). 887 * This key is only supported when _decoding_ content, it cannot 888 * be used to configure an encoder to emit ADTS output. 889 */ 890 public static final String KEY_IS_ADTS = "is-adts"; 891 892 /** 893 * A key describing the channel composition of audio content. This mask 894 * is composed of bits drawn from channel mask definitions in {@link android.media.AudioFormat}. 895 * The associated value is an integer. 896 */ 897 public static final String KEY_CHANNEL_MASK = "channel-mask"; 898 899 /** 900 * A key describing the maximum number of channels that can be output by an audio decoder. 901 * By default, the decoder will output the same number of channels as present in the encoded 902 * stream, if supported. Set this value to limit the number of output channels, and use 903 * the downmix information in the stream, if available. 904 * <p>Values larger than the number of channels in the content to decode behave like the number 905 * of channels in the content (if applicable), for instance passing 99 for a 5.1 audio stream 906 * behaves like passing 6. 907 * <p>This key is only used during decoding. 908 */ 909 public static final String KEY_MAX_OUTPUT_CHANNEL_COUNT = 910 "max-output-channel-count"; 911 912 /** 913 * A key describing the number of frames to trim from the start of the decoded audio stream. 914 * The associated value is an integer. 915 */ 916 public static final String KEY_ENCODER_DELAY = "encoder-delay"; 917 918 /** 919 * A key describing the number of frames to trim from the end of the decoded audio stream. 920 * The associated value is an integer. 921 */ 922 public static final String KEY_ENCODER_PADDING = "encoder-padding"; 923 924 /** 925 * A key describing the AAC profile to be used (AAC audio formats only). 926 * Constants are declared in {@link android.media.MediaCodecInfo.CodecProfileLevel}. 927 */ 928 public static final String KEY_AAC_PROFILE = "aac-profile"; 929 930 /** 931 * A key describing the AAC SBR mode to be used (AAC audio formats only). 932 * The associated value is an integer and can be set to following values: 933 * <ul> 934 * <li>0 - no SBR should be applied</li> 935 * <li>1 - single rate SBR</li> 936 * <li>2 - double rate SBR</li> 937 * </ul> 938 * Note: If this key is not defined the default SRB mode for the desired AAC profile will 939 * be used. 940 * <p>This key is only used during encoding. 941 */ 942 public static final String KEY_AAC_SBR_MODE = "aac-sbr-mode"; 943 944 /** 945 * A key describing the maximum number of channels that can be output by the AAC decoder. 946 * By default, the decoder will output the same number of channels as present in the encoded 947 * stream, if supported. Set this value to limit the number of output channels, and use 948 * the downmix information in the stream, if available. 949 * <p>Values larger than the number of channels in the content to decode behave just 950 * like the actual channel count of the content (e.g. passing 99 for the decoding of 5.1 content 951 * will behave like using 6). 952 * <p>This key is only used during decoding. 953 * @deprecated Use the non-AAC-specific key {@link #KEY_MAX_OUTPUT_CHANNEL_COUNT} instead 954 */ 955 public static final String KEY_AAC_MAX_OUTPUT_CHANNEL_COUNT = "aac-max-output-channel_count"; 956 957 /** 958 * A key describing the Target Reference Level (Target Loudness). 959 * <p>For normalizing loudness across program items, a gain is applied to the audio output so 960 * that the output loudness matches the Target Reference Level. The gain is derived as the 961 * difference between the Target Reference Level and the Program Reference Level (Program 962 * Loudness). The latter can be given in the bitstream and indicates the actual loudness value 963 * of the program item.</p> 964 * <p>The Target Reference Level controls loudness normalization for both MPEG-4 DRC and 965 * MPEG-D DRC. 966 * <p>The value is given as an integer value between 967 * 40 and 127, and is calculated as -4 * Target Reference Level in LKFS. 968 * Therefore, it represents the range of -10 to -31.75 LKFS. 969 * <p>For MPEG-4 DRC, a value of -1 switches off loudness normalization and DRC processing.</p> 970 * <p>For MPEG-D DRC, a value of -1 switches off loudness normalization only. For DRC processing 971 * options of MPEG-D DRC, see {@link #KEY_AAC_DRC_EFFECT_TYPE}</p> 972 * <p>The default value on mobile devices is 64 (-16 LKFS). 973 * <p>This key is only used during decoding. 974 */ 975 public static final String KEY_AAC_DRC_TARGET_REFERENCE_LEVEL = "aac-target-ref-level"; 976 977 /** 978 * A key describing for selecting the DRC effect type for MPEG-D DRC. 979 * The supported values are defined in ISO/IEC 23003-4:2015 and are described as follows: 980 * <table> 981 * <tr><th>Value</th><th>Effect</th></tr> 982 * <tr><th>-1</th><th>Off</th></tr> 983 * <tr><th>0</th><th>None</th></tr> 984 * <tr><th>1</th><th>Late night</th></tr> 985 * <tr><th>2</th><th>Noisy environment</th></tr> 986 * <tr><th>3</th><th>Limited playback range</th></tr> 987 * <tr><th>4</th><th>Low playback level</th></tr> 988 * <tr><th>5</th><th>Dialog enhancement</th></tr> 989 * <tr><th>6</th><th>General compression</th></tr> 990 * </table> 991 * <p>The value -1 (Off) disables DRC processing, while loudness normalization may still be 992 * active and dependent on {@link #KEY_AAC_DRC_TARGET_REFERENCE_LEVEL}.<br> 993 * The value 0 (None) automatically enables DRC processing if necessary to prevent signal 994 * clipping<br> 995 * The value 6 (General compression) can be used for enabling MPEG-D DRC without particular 996 * DRC effect type request.<br> 997 * The default DRC effect type is 3 ("Limited playback range") on mobile devices. 998 * <p>This key is only used during decoding. 999 */ 1000 public static final String KEY_AAC_DRC_EFFECT_TYPE = "aac-drc-effect-type"; 1001 1002 /** 1003 * A key describing the target reference level that was assumed at the encoder for 1004 * calculation of attenuation gains for clipping prevention. 1005 * <p>If it is known, this information can be provided as an integer value between 1006 * 0 and 127, which is calculated as -4 * Encoded Target Level in LKFS. 1007 * If the Encoded Target Level is unknown, the value can be set to -1. 1008 * <p>The default value is -1 (unknown). 1009 * <p>The value is ignored when heavy compression (see {@link #KEY_AAC_DRC_HEAVY_COMPRESSION}) 1010 * or MPEG-D DRC is used. 1011 * <p>This key is only used during decoding. 1012 */ 1013 public static final String KEY_AAC_ENCODED_TARGET_LEVEL = "aac-encoded-target-level"; 1014 1015 /** 1016 * A key describing the boost factor allowing to adapt the dynamics of the output to the 1017 * actual listening requirements. This relies on DRC gain sequences that can be transmitted in 1018 * the encoded bitstream to be able to reduce the dynamics of the output signal upon request. 1019 * This factor enables the user to select how much of the gains are applied. 1020 * <p>Positive gains (boost) and negative gains (attenuation, see 1021 * {@link #KEY_AAC_DRC_ATTENUATION_FACTOR}) can be controlled separately for a better match 1022 * to different use-cases. 1023 * <p>Typically, attenuation gains are sent for loud signal segments, and boost gains are sent 1024 * for soft signal segments. If the output is listened to in a noisy environment, for example, 1025 * the boost factor is used to enable the positive gains, i.e. to amplify soft signal segments 1026 * beyond the noise floor. But for listening late at night, the attenuation 1027 * factor is used to enable the negative gains, to prevent loud signal from surprising 1028 * the listener. In applications which generally need a low dynamic range, both the boost factor 1029 * and the attenuation factor are used in order to enable all DRC gains. 1030 * <p>In order to prevent clipping, it is also recommended to apply the attenuation gains 1031 * in case of a downmix and/or loudness normalization to high target reference levels. 1032 * <p>Both the boost and the attenuation factor parameters are given as integer values 1033 * between 0 and 127, representing the range of the factor of 0 (i.e. don't apply) 1034 * to 1 (i.e. fully apply boost/attenuation gains respectively). 1035 * <p>The default value is 127 (fully apply boost DRC gains). 1036 * <p>This key is only used during decoding. 1037 */ 1038 public static final String KEY_AAC_DRC_BOOST_FACTOR = "aac-drc-boost-level"; 1039 1040 /** 1041 * A key describing the attenuation factor allowing to adapt the dynamics of the output to the 1042 * actual listening requirements. 1043 * See {@link #KEY_AAC_DRC_BOOST_FACTOR} for a description of the role of this attenuation 1044 * factor and the value range. 1045 * <p>The default value is 127 (fully apply attenuation DRC gains). 1046 * <p>This key is only used during decoding. 1047 */ 1048 public static final String KEY_AAC_DRC_ATTENUATION_FACTOR = "aac-drc-cut-level"; 1049 1050 /** 1051 * A key describing the selection of the heavy compression profile for MPEG-4 DRC. 1052 * <p>Two separate DRC gain sequences can be transmitted in one bitstream: light compression 1053 * and heavy compression. When selecting the application of the heavy compression, one of 1054 * the sequences is selected: 1055 * <ul> 1056 * <li>0 enables light compression,</li> 1057 * <li>1 enables heavy compression instead. 1058 * </ul> 1059 * Note that heavy compression doesn't offer the features of scaling of DRC gains 1060 * (see {@link #KEY_AAC_DRC_BOOST_FACTOR} and {@link #KEY_AAC_DRC_ATTENUATION_FACTOR} for the 1061 * boost and attenuation factors), and frequency-selective (multiband) DRC. 1062 * Light compression usually contains clipping prevention for stereo downmixing while heavy 1063 * compression, if additionally provided in the bitstream, is usually stronger, and contains 1064 * clipping prevention for stereo and mono downmixing. 1065 * <p>The default is 1 (heavy compression). 1066 * <p>This key is only used during decoding. 1067 */ 1068 public static final String KEY_AAC_DRC_HEAVY_COMPRESSION = "aac-drc-heavy-compression"; 1069 1070 /** 1071 * A key to retrieve the output loudness of a decoded bitstream. 1072 * <p>If loudness normalization is active, the value corresponds to the Target Reference Level 1073 * (see {@link #KEY_AAC_DRC_TARGET_REFERENCE_LEVEL}).<br> 1074 * If loudness normalization is not active, the value corresponds to the loudness metadata 1075 * given in the bitstream. 1076 * <p>The value is retrieved with getInteger() and is given as an integer value between 0 and 1077 * 231. It is calculated as -4 * Output Loudness in LKFS. Therefore, it represents the range of 1078 * 0 to -57.75 LKFS. 1079 * <p>A value of -1 indicates that no loudness metadata is present in the bitstream. 1080 * <p>Loudness metadata can originate from MPEG-4 DRC or MPEG-D DRC. 1081 * <p>This key is only used during decoding. 1082 */ 1083 public static final String KEY_AAC_DRC_OUTPUT_LOUDNESS = "aac-drc-output-loudness"; 1084 1085 /** 1086 * A key describing the album mode for MPEG-D DRC as defined in ISO/IEC 23003-4. 1087 * <p>The associated value is an integer and can be set to following values: 1088 * <table> 1089 * <tr><th>Value</th><th>Album Mode</th></tr> 1090 * <tr><th>0</th><th>disabled</th></tr> 1091 * <tr><th>1</th><th>enabled</th></tr> 1092 * </table> 1093 * <p>Disabled album mode leads to application of gain sequences for fading in and out, if 1094 * provided in the bitstream. Enabled album mode makes use of dedicated album loudness 1095 * information, if provided in the bitstream. 1096 * <p>The default value is 0 (album mode disabled). 1097 * <p>This key is only used during decoding. 1098 */ 1099 public static final String KEY_AAC_DRC_ALBUM_MODE = "aac-drc-album-mode"; 1100 1101 /** 1102 * A key describing the FLAC compression level to be used (FLAC audio format only). 1103 * The associated value is an integer ranging from 0 (fastest, least compression) 1104 * to 8 (slowest, most compression). 1105 */ 1106 public static final String KEY_FLAC_COMPRESSION_LEVEL = "flac-compression-level"; 1107 1108 /** 1109 * A key describing the MPEG-H stream profile-level indication. 1110 * 1111 * See ISO_IEC_23008-3;2019 MHADecoderConfigurationRecord mpegh3daProfileLevelIndication. 1112 */ 1113 public static final String KEY_MPEGH_PROFILE_LEVEL_INDICATION = 1114 "mpegh-profile-level-indication"; 1115 1116 /** 1117 * A key describing the MPEG-H stream compatible sets. 1118 * 1119 * See FDAmd_2 of ISO_IEC_23008-3;2019 MHAProfileAndLevelCompatibilitySetBox. 1120 */ 1121 public static final String KEY_MPEGH_COMPATIBLE_SETS = "mpegh-compatible-sets"; 1122 1123 /** 1124 * A key describing the MPEG-H stream reference channel layout. 1125 * 1126 * See ISO_IEC_23008-3;2019 MHADecoderConfigurationRecord referenceChannelLayout 1127 * and ISO_IEC_23001‐8 ChannelConfiguration value. 1128 */ 1129 public static final String KEY_MPEGH_REFERENCE_CHANNEL_LAYOUT = 1130 "mpegh-reference-channel-layout"; 1131 1132 /** 1133 * A key describing the encoding complexity. 1134 * The associated value is an integer. These values are device and codec specific, 1135 * but lower values generally result in faster and/or less power-hungry encoding. 1136 * 1137 * @see MediaCodecInfo.EncoderCapabilities#getComplexityRange() 1138 */ 1139 public static final String KEY_COMPLEXITY = "complexity"; 1140 1141 /** 1142 * A key describing the desired encoding quality. 1143 * The associated value is an integer. This key is only supported for encoders 1144 * that are configured in constant-quality mode. These values are device and 1145 * codec specific, but lower values generally result in more efficient 1146 * (smaller-sized) encoding. 1147 * 1148 * @see MediaCodecInfo.EncoderCapabilities#getQualityRange() 1149 */ 1150 public static final String KEY_QUALITY = "quality"; 1151 1152 /** 1153 * A key describing the desired codec priority. 1154 * <p> 1155 * The associated value is an integer. Higher value means lower priority. 1156 * <p> 1157 * Currently, only two levels are supported:<br> 1158 * 0: realtime priority - meaning that the codec shall support the given 1159 * performance configuration (e.g. framerate) at realtime. This should 1160 * only be used by media playback, capture, and possibly by realtime 1161 * communication scenarios if best effort performance is not suitable.<br> 1162 * 1: non-realtime priority (best effort). 1163 * <p> 1164 * This is a hint used at codec configuration and resource planning - to understand 1165 * the realtime requirements of the application; however, due to the nature of 1166 * media components, performance is not guaranteed. 1167 * 1168 */ 1169 public static final String KEY_PRIORITY = "priority"; 1170 1171 /** 1172 * A key describing the desired operating frame rate for video or sample rate for audio 1173 * that the codec will need to operate at. 1174 * <p> 1175 * The associated value is an integer or a float representing frames-per-second or 1176 * samples-per-second 1177 * <p> 1178 * This is used for cases like high-speed/slow-motion video capture, where the video encoder 1179 * format contains the target playback rate (e.g. 30fps), but the component must be able to 1180 * handle the high operating capture rate (e.g. 240fps). 1181 * <p> 1182 * This rate will be used by codec for resource planning and setting the operating points. 1183 * 1184 */ 1185 public static final String KEY_OPERATING_RATE = "operating-rate"; 1186 1187 /** 1188 * A key describing the desired profile to be used by an encoder. 1189 * <p> 1190 * The associated value is an integer. 1191 * Constants are declared in {@link MediaCodecInfo.CodecProfileLevel}. 1192 * This key is used as a hint, and is only supported for codecs 1193 * that specify a profile. When configuring profile, encoder configuration 1194 * may fail if other parameters are not compatible with the desired 1195 * profile or if the desired profile is not supported, but it may also 1196 * fail silently (where the encoder ends up using a different, compatible profile.) 1197 * <p> 1198 * It is recommended that the profile is set for all encoders. For more information, see 1199 * the <i>Encoder Profiles</i> section of the {@link MediaCodec} API reference. 1200 * <p class="note"> 1201 * <strong>Note:</strong> Codecs are free to use all the available 1202 * coding tools at the specified profile, but may ultimately choose to not do so. 1203 * <p class="note"> 1204 * <strong>Note:</strong> When configuring video encoders, profile (if set) must be 1205 * set together with {@link #KEY_LEVEL level}. 1206 * 1207 * @see MediaCodecInfo.CodecCapabilities#profileLevels 1208 */ 1209 public static final String KEY_PROFILE = "profile"; 1210 1211 /** 1212 * A key describing the desired profile to be used by an encoder. 1213 * <p> 1214 * The associated value is an integer. 1215 * Constants are declared in {@link MediaCodecInfo.CodecProfileLevel}. 1216 * This key is used as a further hint when specifying a desired profile, 1217 * and is only supported for codecs that specify a level. 1218 * <p> 1219 * This key is ignored if the {@link #KEY_PROFILE profile} is not specified. 1220 * Otherwise, the value should be a level compatible with the configured encoding 1221 * parameters. 1222 * <p class="note"> 1223 * <strong>Note:</strong> This key cannot be used to constrain the encoder's 1224 * output to a maximum encoding level. Encoders are free to target a different 1225 * level if the configured encoding parameters dictate it. Nevertheless, 1226 * encoders shall use (and encode) a level sufficient to decode the generated 1227 * bitstream, though they may exceed the (Video) Buffering Verifier limits for 1228 * that encoded level. 1229 * 1230 * @see MediaCodecInfo.CodecCapabilities#profileLevels 1231 */ 1232 public static final String KEY_LEVEL = "level"; 1233 1234 /** 1235 * An optional key describing the desired encoder latency in frames. This is an optional 1236 * parameter that applies only to video encoders. If encoder supports it, it should ouput 1237 * at least one output frame after being queued the specified number of frames. This key 1238 * is ignored if the video encoder does not support the latency feature. Use the output 1239 * format to verify that this feature was enabled and the actual value used by the encoder. 1240 * <p> 1241 * If the key is not specified, the default latency will be implenmentation specific. 1242 * The associated value is an integer. 1243 */ 1244 public static final String KEY_LATENCY = "latency"; 1245 1246 /** 1247 * An optional key describing the maximum number of non-display-order coded frames. 1248 * This is an optional parameter that applies only to video encoders. Application should 1249 * check the value for this key in the output format to see if codec will produce 1250 * non-display-order coded frames. If encoder supports it, the output frames' order will be 1251 * different from the display order and each frame's display order could be retrived from 1252 * {@link MediaCodec.BufferInfo#presentationTimeUs}. Before API level 27, application may 1253 * receive non-display-order coded frames even though the application did not request it. 1254 * Note: Application should not rearrange the frames to display order before feeding them 1255 * to {@link MediaMuxer#writeSampleData}. 1256 * <p> 1257 * The default value is 0. 1258 */ 1259 public static final String KEY_OUTPUT_REORDER_DEPTH = "output-reorder-depth"; 1260 1261 /** 1262 * A key describing the desired clockwise rotation on an output surface. 1263 * This key is only used when the codec is configured using an output surface. 1264 * The associated value is an integer, representing degrees. Supported values 1265 * are 0, 90, 180 or 270. This is an optional field; if not specified, rotation 1266 * defaults to 0. 1267 * 1268 * @see MediaCodecInfo.CodecCapabilities#profileLevels 1269 */ 1270 public static final String KEY_ROTATION = "rotation-degrees"; 1271 1272 /** 1273 * A key describing the desired bitrate mode to be used by an encoder. 1274 * Constants are declared in {@link MediaCodecInfo.EncoderCapabilities}. 1275 * 1276 * @see MediaCodecInfo.EncoderCapabilities#isBitrateModeSupported(int) 1277 */ 1278 public static final String KEY_BITRATE_MODE = "bitrate-mode"; 1279 1280 /** 1281 * A key describing the maximum Quantization Parameter allowed for encoding video. 1282 * This key applies to all three video picture types (I, P, and B). 1283 * The value is used directly for picture type I; a per-mime formula is used 1284 * to calculate the value for the remaining picture types. 1285 * 1286 * This calculation can be avoided by directly specifying values for each picture type 1287 * using the type-specific keys {@link #KEY_VIDEO_QP_I_MAX}, {@link #KEY_VIDEO_QP_P_MAX}, 1288 * and {@link #KEY_VIDEO_QP_B_MAX}. 1289 * 1290 * The associated value is an integer. 1291 */ 1292 public static final String KEY_VIDEO_QP_MAX = "video-qp-max"; 1293 1294 /** 1295 * A key describing the minimum Quantization Parameter allowed for encoding video. 1296 * This key applies to all three video frame types (I, P, and B). 1297 * The value is used directly for picture type I; a per-mime formula is used 1298 * to calculate the value for the remaining picture types. 1299 * 1300 * This calculation can be avoided by directly specifying values for each picture type 1301 * using the type-specific keys {@link #KEY_VIDEO_QP_I_MIN}, {@link #KEY_VIDEO_QP_P_MIN}, 1302 * and {@link #KEY_VIDEO_QP_B_MIN}. 1303 * 1304 * The associated value is an integer. 1305 */ 1306 public static final String KEY_VIDEO_QP_MIN = "video-qp-min"; 1307 1308 /** 1309 * A key describing the maximum Quantization Parameter allowed for encoding video. 1310 * This value applies to video I-frames. 1311 * 1312 * The associated value is an integer. 1313 */ 1314 public static final String KEY_VIDEO_QP_I_MAX = "video-qp-i-max"; 1315 1316 /** 1317 * A key describing the minimum Quantization Parameter allowed for encoding video. 1318 * This value applies to video I-frames. 1319 * 1320 * The associated value is an integer. 1321 */ 1322 public static final String KEY_VIDEO_QP_I_MIN = "video-qp-i-min"; 1323 1324 /** 1325 * A key describing the maximum Quantization Parameter allowed for encoding video. 1326 * This value applies to video P-frames. 1327 * 1328 * The associated value is an integer. 1329 */ 1330 public static final String KEY_VIDEO_QP_P_MAX = "video-qp-p-max"; 1331 1332 /** 1333 * A key describing the minimum Quantization Parameter allowed for encoding video. 1334 * This value applies to video P-frames. 1335 * 1336 * The associated value is an integer. 1337 */ 1338 public static final String KEY_VIDEO_QP_P_MIN = "video-qp-p-min"; 1339 1340 /** 1341 * A key describing the maximum Quantization Parameter allowed for encoding video. 1342 * This value applies to video B-frames. 1343 * 1344 * The associated value is an integer. 1345 */ 1346 public static final String KEY_VIDEO_QP_B_MAX = "video-qp-b-max"; 1347 1348 /** 1349 * A key describing the minimum Quantization Parameter allowed for encoding video. 1350 * This value applies to video B-frames. 1351 * 1352 * The associated value is an integer. 1353 */ 1354 public static final String KEY_VIDEO_QP_B_MIN = "video-qp-b-min"; 1355 1356 /** 1357 * A key describing the level of encoding statistics information emitted from video encoder. 1358 * 1359 * The associated value is an integer. 1360 */ 1361 public static final String KEY_VIDEO_ENCODING_STATISTICS_LEVEL = 1362 "video-encoding-statistics-level"; 1363 1364 /** 1365 * Encoding Statistics Level None. 1366 * Encoder generates no information about Encoding statistics. 1367 */ 1368 public static final int VIDEO_ENCODING_STATISTICS_LEVEL_NONE = 0; 1369 1370 /** 1371 * Encoding Statistics Level 1. 1372 * Encoder generates {@link MediaFormat#KEY_PICTURE_TYPE} and 1373 * {@link MediaFormat#KEY_VIDEO_QP_AVERAGE} for each frame. 1374 */ 1375 public static final int VIDEO_ENCODING_STATISTICS_LEVEL_1 = 1; 1376 1377 /** @hide */ 1378 @IntDef({ 1379 VIDEO_ENCODING_STATISTICS_LEVEL_NONE, 1380 VIDEO_ENCODING_STATISTICS_LEVEL_1, 1381 }) 1382 @Retention(RetentionPolicy.SOURCE) 1383 public @interface VideoEncodingStatisticsLevel {} 1384 1385 /** 1386 * A key describing the per-frame average block QP (Quantization Parameter). 1387 * This is a part of a video 'Encoding Statistics' export feature. 1388 * This value is emitted from video encoder for a video frame. 1389 * The average value is rounded to the nearest integer value. 1390 * 1391 * The associated value is an integer. 1392 */ 1393 public static final String KEY_VIDEO_QP_AVERAGE = "video-qp-average"; 1394 1395 /** 1396 * A key describing the picture type of the encoded frame. 1397 * This is a part of a video 'Encoding Statistics' export feature. 1398 * This value is emitted from video encoder for a video frame. 1399 * 1400 * The associated value is an integer. 1401 */ 1402 public static final String KEY_PICTURE_TYPE = "picture-type"; 1403 1404 /** Picture Type is unknown. */ 1405 public static final int PICTURE_TYPE_UNKNOWN = 0; 1406 1407 /** Picture Type is I Frame. */ 1408 public static final int PICTURE_TYPE_I = 1; 1409 1410 /** Picture Type is P Frame. */ 1411 public static final int PICTURE_TYPE_P = 2; 1412 1413 /** Picture Type is B Frame. */ 1414 public static final int PICTURE_TYPE_B = 3; 1415 1416 /** @hide */ 1417 @IntDef({ 1418 PICTURE_TYPE_UNKNOWN, 1419 PICTURE_TYPE_I, 1420 PICTURE_TYPE_P, 1421 PICTURE_TYPE_B, 1422 }) 1423 @Retention(RetentionPolicy.SOURCE) 1424 public @interface PictureType {} 1425 1426 /** 1427 * A key describing the audio session ID of the AudioTrack associated 1428 * to a tunneled video codec. 1429 * The associated value is an integer. 1430 * 1431 * @see MediaCodecInfo.CodecCapabilities#FEATURE_TunneledPlayback 1432 */ 1433 public static final String KEY_AUDIO_SESSION_ID = "audio-session-id"; 1434 1435 /** 1436 * A key describing the audio hardware sync ID of the AudioTrack associated 1437 * to a tunneled video codec. The associated value is an integer. 1438 * 1439 * @hide 1440 * 1441 * @see MediaCodecInfo.CodecCapabilities#FEATURE_TunneledPlayback 1442 * @see AudioManager#getAudioHwSyncForSession 1443 */ 1444 public static final String KEY_AUDIO_HW_SYNC = "audio-hw-sync"; 1445 1446 /** 1447 * A key for boolean AUTOSELECT behavior for the track. Tracks with AUTOSELECT=true 1448 * are considered when automatically selecting a track without specific user 1449 * choice, based on the current locale. 1450 * This is currently only used for subtitle tracks, when the user selected 1451 * 'Default' for the captioning locale. 1452 * The associated value is an integer, where non-0 means TRUE. This is an optional 1453 * field; if not specified, AUTOSELECT defaults to TRUE. 1454 */ 1455 public static final String KEY_IS_AUTOSELECT = "is-autoselect"; 1456 1457 /** 1458 * A key for boolean DEFAULT behavior for the track. The track with DEFAULT=true is 1459 * selected in the absence of a specific user choice. 1460 * This is currently used in two scenarios: 1461 * 1) for subtitle tracks, when the user selected 'Default' for the captioning locale. 1462 * 2) for a {@link #MIMETYPE_IMAGE_ANDROID_HEIC} / {@link #MIMETYPE_IMAGE_AVIF} track, 1463 * indicating the image is the primary item in the file. 1464 * 1465 * The associated value is an integer, where non-0 means TRUE. This is an optional 1466 * field; if not specified, DEFAULT is considered to be FALSE. 1467 */ 1468 public static final String KEY_IS_DEFAULT = "is-default"; 1469 1470 /** 1471 * A key for the FORCED field for subtitle tracks. True if it is a 1472 * forced subtitle track. Forced subtitle tracks are essential for the 1473 * content and are shown even when the user turns off Captions. They 1474 * are used for example to translate foreign/alien dialogs or signs. 1475 * The associated value is an integer, where non-0 means TRUE. This is an 1476 * optional field; if not specified, FORCED defaults to FALSE. 1477 */ 1478 public static final String KEY_IS_FORCED_SUBTITLE = "is-forced-subtitle"; 1479 1480 /** 1481 * A key describing the number of haptic channels in an audio format. 1482 * The associated value is an integer. 1483 */ 1484 public static final String KEY_HAPTIC_CHANNEL_COUNT = "haptic-channel-count"; 1485 1486 /** @hide */ 1487 public static final String KEY_IS_TIMED_TEXT = "is-timed-text"; 1488 1489 // The following color aspect values must be in sync with the ones in HardwareAPI.h. 1490 /** 1491 * An optional key describing the color primaries, white point and 1492 * luminance factors for video content. 1493 * 1494 * The associated value is an integer: 0 if unspecified, or one of the 1495 * COLOR_STANDARD_ values. 1496 */ 1497 public static final String KEY_COLOR_STANDARD = "color-standard"; 1498 1499 /** BT.709 color chromacity coordinates with KR = 0.2126, KB = 0.0722. */ 1500 public static final int COLOR_STANDARD_BT709 = 1; 1501 1502 /** BT.601 625 color chromacity coordinates with KR = 0.299, KB = 0.114. */ 1503 public static final int COLOR_STANDARD_BT601_PAL = 2; 1504 1505 /** BT.601 525 color chromacity coordinates with KR = 0.299, KB = 0.114. */ 1506 public static final int COLOR_STANDARD_BT601_NTSC = 4; 1507 1508 /** BT.2020 color chromacity coordinates with KR = 0.2627, KB = 0.0593. */ 1509 public static final int COLOR_STANDARD_BT2020 = 6; 1510 1511 /** @hide */ 1512 @IntDef({ 1513 COLOR_STANDARD_BT709, 1514 COLOR_STANDARD_BT601_PAL, 1515 COLOR_STANDARD_BT601_NTSC, 1516 COLOR_STANDARD_BT2020, 1517 }) 1518 @Retention(RetentionPolicy.SOURCE) 1519 public @interface ColorStandard {} 1520 1521 /** 1522 * An optional key describing the opto-electronic transfer function used 1523 * for the video content. 1524 * 1525 * The associated value is an integer: 0 if unspecified, or one of the 1526 * COLOR_TRANSFER_ values. 1527 */ 1528 public static final String KEY_COLOR_TRANSFER = "color-transfer"; 1529 1530 /** Linear transfer characteristic curve. */ 1531 public static final int COLOR_TRANSFER_LINEAR = 1; 1532 1533 /** SMPTE 170M transfer characteristic curve used by BT.601/BT.709/BT.2020. This is the curve 1534 * used by most non-HDR video content. */ 1535 public static final int COLOR_TRANSFER_SDR_VIDEO = 3; 1536 1537 /** SMPTE ST 2084 transfer function. This is used by some HDR video content. */ 1538 public static final int COLOR_TRANSFER_ST2084 = 6; 1539 1540 /** ARIB STD-B67 hybrid-log-gamma transfer function. This is used by some HDR video content. */ 1541 public static final int COLOR_TRANSFER_HLG = 7; 1542 1543 /** @hide */ 1544 @IntDef({ 1545 COLOR_TRANSFER_LINEAR, 1546 COLOR_TRANSFER_SDR_VIDEO, 1547 COLOR_TRANSFER_ST2084, 1548 COLOR_TRANSFER_HLG, 1549 }) 1550 @Retention(RetentionPolicy.SOURCE) 1551 public @interface ColorTransfer {} 1552 1553 /** 1554 * An optional key describing the range of the component values of the video content. 1555 * 1556 * The associated value is an integer: 0 if unspecified, or one of the 1557 * COLOR_RANGE_ values. 1558 */ 1559 public static final String KEY_COLOR_RANGE = "color-range"; 1560 1561 /** Limited range. Y component values range from 16 to 235 for 8-bit content. 1562 * Cr, Cy values range from 16 to 240 for 8-bit content. 1563 * This is the default for video content. */ 1564 public static final int COLOR_RANGE_LIMITED = 2; 1565 1566 /** Full range. Y, Cr and Cb component values range from 0 to 255 for 8-bit content. */ 1567 public static final int COLOR_RANGE_FULL = 1; 1568 1569 /** @hide */ 1570 @IntDef({ 1571 COLOR_RANGE_LIMITED, 1572 COLOR_RANGE_FULL, 1573 }) 1574 @Retention(RetentionPolicy.SOURCE) 1575 public @interface ColorRange {} 1576 1577 /** 1578 * An optional key describing the static metadata of HDR (high-dynamic-range) video content. 1579 * 1580 * The associated value is a ByteBuffer. This buffer contains the raw contents of the 1581 * Static Metadata Descriptor (including the descriptor ID) of an HDMI Dynamic Range and 1582 * Mastering InfoFrame as defined by CTA-861.3. This key must be provided to video decoders 1583 * for HDR video content unless this information is contained in the bitstream and the video 1584 * decoder supports an HDR-capable profile. This key must be provided to video encoders for 1585 * HDR video content. 1586 */ 1587 public static final String KEY_HDR_STATIC_INFO = "hdr-static-info"; 1588 1589 /** 1590 * An optional key describing the HDR10+ metadata of the video content. 1591 * 1592 * The associated value is a ByteBuffer containing HDR10+ metadata conforming to the 1593 * user_data_registered_itu_t_t35() syntax of SEI message for ST 2094-40. This key will 1594 * be present on: 1595 *<p> 1596 * - The formats of output buffers of a decoder configured for HDR10+ profiles (such as 1597 * {@link MediaCodecInfo.CodecProfileLevel#VP9Profile2HDR10Plus}, {@link 1598 * MediaCodecInfo.CodecProfileLevel#VP9Profile3HDR10Plus} or {@link 1599 * MediaCodecInfo.CodecProfileLevel#HEVCProfileMain10HDR10Plus}), or 1600 *<p> 1601 * - The formats of output buffers of an encoder configured for an HDR10+ profiles that 1602 * uses out-of-band metadata (such as {@link 1603 * MediaCodecInfo.CodecProfileLevel#VP9Profile2HDR10Plus} or {@link 1604 * MediaCodecInfo.CodecProfileLevel#VP9Profile3HDR10Plus}). 1605 * 1606 * @see MediaCodec#PARAMETER_KEY_HDR10_PLUS_INFO 1607 */ 1608 public static final String KEY_HDR10_PLUS_INFO = "hdr10-plus-info"; 1609 1610 /** 1611 * An optional key describing the opto-electronic transfer function 1612 * requested for the output video content. 1613 * 1614 * The associated value is an integer: 0 if unspecified, or one of the 1615 * COLOR_TRANSFER_ values. When unspecified the component will not touch the 1616 * video content; otherwise the component will tone-map the raw video frame 1617 * to match the requested transfer function. 1618 * 1619 * After configure, component's input format will contain this key to note 1620 * whether the request is supported or not. If the value in the input format 1621 * is the same as the requested value, the request is supported. The value 1622 * is set to 0 if unsupported. 1623 */ 1624 public static final String KEY_COLOR_TRANSFER_REQUEST = "color-transfer-request"; 1625 1626 /** 1627 * A key describing a unique ID for the content of a media track. 1628 * 1629 * <p>This key is used by {@link MediaExtractor}. Some extractors provide multiple encodings 1630 * of the same track (e.g. float audio tracks for FLAC and WAV may be expressed as two 1631 * tracks via MediaExtractor: a normal PCM track for backward compatibility, and a float PCM 1632 * track for added fidelity. Similarly, Dolby Vision extractor may provide a baseline SDR 1633 * version of a DV track.) This key can be used to identify which MediaExtractor tracks refer 1634 * to the same underlying content. 1635 * </p> 1636 * 1637 * The associated value is an integer. 1638 */ 1639 public static final String KEY_TRACK_ID = "track-id"; 1640 1641 /** 1642 * A key describing the system id of the conditional access system used to scramble 1643 * a media track. 1644 * <p> 1645 * This key is set by {@link MediaExtractor} if the track is scrambled with a conditional 1646 * access system, regardless of the presence of a valid {@link MediaCas} object. 1647 * <p> 1648 * The associated value is an integer. 1649 * @hide 1650 */ 1651 public static final String KEY_CA_SYSTEM_ID = "ca-system-id"; 1652 1653 /** 1654 * A key describing the {@link MediaCas.Session} object associated with a media track. 1655 * <p> 1656 * This key is set by {@link MediaExtractor} if the track is scrambled with a conditional 1657 * access system, after it receives a valid {@link MediaCas} object. 1658 * <p> 1659 * The associated value is a ByteBuffer. 1660 * @hide 1661 */ 1662 public static final String KEY_CA_SESSION_ID = "ca-session-id"; 1663 1664 /** 1665 * A key describing the private data in the CA_descriptor associated with a media track. 1666 * <p> 1667 * This key is set by {@link MediaExtractor} if the track is scrambled with a conditional 1668 * access system, before it receives a valid {@link MediaCas} object. 1669 * <p> 1670 * The associated value is a ByteBuffer. 1671 * @hide 1672 */ 1673 public static final String KEY_CA_PRIVATE_DATA = "ca-private-data"; 1674 1675 /** 1676 * A key describing the maximum number of B frames between I or P frames, 1677 * to be used by a video encoder. 1678 * The associated value is an integer. The default value is 0, which means 1679 * that no B frames are allowed. Note that non-zero value does not guarantee 1680 * B frames; it's up to the encoder to decide. 1681 */ 1682 public static final String KEY_MAX_B_FRAMES = "max-bframes"; 1683 1684 /** 1685 * A key for applications to opt out of allowing 1686 * a Surface to discard undisplayed/unconsumed frames 1687 * as means to catch up after falling behind. 1688 * This value is an integer. 1689 * The value 0 indicates the surface is not allowed to drop frames. 1690 * The value 1 indicates the surface is allowed to drop frames. 1691 * 1692 * {@link MediaCodec} describes the semantics. 1693 */ 1694 public static final String KEY_ALLOW_FRAME_DROP = "allow-frame-drop"; 1695 1696 /** 1697 * A key describing the desired codec importance for the application. 1698 * <p> 1699 * The associated value is a positive integer including zero. 1700 * Higher value means lesser importance. 1701 * <p> 1702 * The resource manager may use the codec importance, along with other factors 1703 * when reclaiming codecs from an application. 1704 * The specifics of reclaim policy is device dependent, but specifying the codec importance, 1705 * will allow the resource manager to prioritize reclaiming less important codecs 1706 * (assigned higher values) from the (reclaim) requesting application first. 1707 * So, the codec importance is only relevant within the context of that application. 1708 * <p> 1709 * The codec importance can be set: 1710 * <ul> 1711 * <li>through {@link MediaCodec#configure}. </li> 1712 * <li>through {@link MediaCodec#setParameters} if the codec has been configured already, 1713 * which allows the users to change the codec importance multiple times. 1714 * </ul> 1715 * Any change/update in codec importance is guaranteed upon the completion of the function call 1716 * that sets the codec importance. So, in case of concurrent codec operations, 1717 * make sure to wait for the change in codec importance, before using another codec. 1718 * Note that unless specified, by default the codecs will have highest importance (of value 0). 1719 * 1720 */ 1721 @FlaggedApi(FLAG_CODEC_IMPORTANCE) 1722 public static final String KEY_IMPORTANCE = "importance"; 1723 1724 /** @hide */ 1725 @IntDef(flag = true, prefix = {"FLAG_SECURITY_MODEL_"}, value = { 1726 FLAG_SECURITY_MODEL_SANDBOXED, 1727 FLAG_SECURITY_MODEL_MEMORY_SAFE, 1728 FLAG_SECURITY_MODEL_TRUSTED_CONTENT_ONLY, 1729 }) 1730 @Retention(RetentionPolicy.SOURCE) 1731 public @interface SecurityModelFlag {} 1732 1733 /** 1734 * Flag for {@link MediaCodecInfo#SECURITY_MODEL_SANDBOXED}. 1735 */ 1736 @FlaggedApi(FLAG_IN_PROCESS_SW_AUDIO_CODEC) 1737 public static final int FLAG_SECURITY_MODEL_SANDBOXED = 1738 (1 << MediaCodecInfo.SECURITY_MODEL_SANDBOXED); 1739 /** 1740 * Flag for {@link MediaCodecInfo#SECURITY_MODEL_MEMORY_SAFE}. 1741 */ 1742 @FlaggedApi(FLAG_IN_PROCESS_SW_AUDIO_CODEC) 1743 public static final int FLAG_SECURITY_MODEL_MEMORY_SAFE = 1744 (1 << MediaCodecInfo.SECURITY_MODEL_MEMORY_SAFE); 1745 /** 1746 * Flag for {@link MediaCodecInfo#SECURITY_MODEL_TRUSTED_CONTENT_ONLY}. 1747 */ 1748 @FlaggedApi(FLAG_IN_PROCESS_SW_AUDIO_CODEC) 1749 public static final int FLAG_SECURITY_MODEL_TRUSTED_CONTENT_ONLY = 1750 (1 << MediaCodecInfo.SECURITY_MODEL_TRUSTED_CONTENT_ONLY); 1751 1752 /** 1753 * A key describing the requested security model as flags. 1754 * <p> 1755 * The associated value is a flag of the following values: 1756 * {@link FLAG_SECURITY_MODEL_SANDBOXED}, 1757 * {@link FLAG_SECURITY_MODEL_MEMORY_SAFE}, 1758 * {@link FLAG_SECURITY_MODEL_TRUSTED_CONTENT_ONLY}. The default value is 1759 * {@link FLAG_SECURITY_MODEL_SANDBOXED}. 1760 * <p> 1761 * When passed to {@link MediaCodecList#findDecoderForFormat} or 1762 * {@link MediaCodecList#findEncoderForFormat}, MediaCodecList filters 1763 * the security model of the codecs according to this flag value. 1764 * <p> 1765 * When passed to {@link MediaCodec#configure}, MediaCodec verifies 1766 * the security model matches the flag value passed, and throws 1767 * {@link java.lang.IllegalArgumentException} if the model does not match. 1768 * <p> 1769 * @see MediaCodecInfo#getSecurityModel 1770 * @see MediaCodecList#findDecoderForFormat 1771 * @see MediaCodecList#findEncoderForFormat 1772 */ 1773 @FlaggedApi(FLAG_IN_PROCESS_SW_AUDIO_CODEC) 1774 public static final String KEY_SECURITY_MODEL = "security-model"; 1775 1776 /** 1777 * QpOffsetRect constitutes the metadata required for encoding a region of interest in an 1778 * image or a video frame. The region of interest is represented by a rectangle. The four 1779 * integer coordinates of the rectangle are stored in fields left, top, right, bottom. 1780 * Note that the right and bottom coordinates are exclusive. 1781 * This is paired with a suggestive qp offset information that is to be used during encoding 1782 * of the blocks belonging to the to the box. 1783 */ 1784 @FlaggedApi(FLAG_REGION_OF_INTEREST) 1785 public static final class QpOffsetRect { 1786 private Rect mContour; 1787 private int mQpOffset; 1788 1789 /** 1790 * Create a new region of interest with the specified coordinates and qpOffset. Note: no 1791 * range checking is performed, so the caller must ensure that left >= 0, left <= right, 1792 * top >= 0 and top <= bottom. Note that the right and bottom coordinates are exclusive. 1793 * 1794 * @param contour Rectangle specifying the region of interest 1795 * @param qpOffset qpOffset to be used for the blocks in the specified rectangle 1796 */ QpOffsetRect(@onNull Rect contour, int qpOffset)1797 public QpOffsetRect(@NonNull Rect contour, int qpOffset) { 1798 mContour = contour; 1799 mQpOffset = qpOffset; 1800 } 1801 1802 /** 1803 * Update the region of interest information with the specified coordinates and qpOffset 1804 * 1805 * @param contour Rectangle specifying the region of interest 1806 * @param qpOffset qpOffset to be used for the blocks in the specified rectangle 1807 */ set(@onNull Rect contour, int qpOffset)1808 public void set(@NonNull Rect contour, int qpOffset) { 1809 mContour = contour; 1810 mQpOffset = qpOffset; 1811 } 1812 1813 /** 1814 * @return Return a string representation of qpOffsetRect in a compact form. 1815 * Helper function to insert key {@link #PARAMETER_KEY_QP_OFFSET_RECTS} in MediaFormat 1816 */ 1817 @NonNull flattenToString()1818 public String flattenToString() { 1819 return TextUtils.formatSimple("%d,%d-%d,%d=%d;", mContour.top, mContour.left, 1820 mContour.bottom, mContour.right, mQpOffset); 1821 } 1822 1823 /** 1824 * @return Return a string representation of qpOffsetRect in a compact form. 1825 * Helper function to insert key {@link #PARAMETER_KEY_QP_OFFSET_RECTS} in MediaFormat 1826 */ 1827 @NonNull flattenToString(@onNull List<QpOffsetRect> qpOffsetRects)1828 public static String flattenToString(@NonNull List<QpOffsetRect> qpOffsetRects) { 1829 StringBuilder builder = new StringBuilder(); 1830 for (QpOffsetRect qpOffsetRect : qpOffsetRects) { 1831 builder.append(qpOffsetRect.flattenToString()); 1832 } 1833 return builder.toString(); 1834 } 1835 } 1836 MediaFormat(@onNull Map<String, Object> map)1837 /* package private */ MediaFormat(@NonNull Map<String, Object> map) { 1838 mMap = map; 1839 } 1840 1841 /** 1842 * Creates an empty MediaFormat 1843 */ MediaFormat()1844 public MediaFormat() { 1845 mMap = new HashMap(); 1846 } 1847 1848 @UnsupportedAppUsage getMap()1849 /* package private */ Map<String, Object> getMap() { 1850 return mMap; 1851 } 1852 1853 /** 1854 * Returns true iff a key of the given name exists in the format. 1855 */ containsKey(@onNull String name)1856 public final boolean containsKey(@NonNull String name) { 1857 return mMap.containsKey(name); 1858 } 1859 1860 /** 1861 * Returns true iff a feature of the given name exists in the format. 1862 */ containsFeature(@onNull String name)1863 public final boolean containsFeature(@NonNull String name) { 1864 return mMap.containsKey(KEY_FEATURE_ + name); 1865 } 1866 1867 public static final int TYPE_NULL = 0; 1868 public static final int TYPE_INTEGER = 1; 1869 public static final int TYPE_LONG = 2; 1870 public static final int TYPE_FLOAT = 3; 1871 public static final int TYPE_STRING = 4; 1872 public static final int TYPE_BYTE_BUFFER = 5; 1873 1874 /** @hide */ 1875 @IntDef({ 1876 TYPE_NULL, 1877 TYPE_INTEGER, 1878 TYPE_LONG, 1879 TYPE_FLOAT, 1880 TYPE_STRING, 1881 TYPE_BYTE_BUFFER 1882 }) 1883 @Retention(RetentionPolicy.SOURCE) 1884 public @interface Type {} 1885 1886 /** 1887 * Returns the value type for a key. If the key does not exist, it returns TYPE_NULL. 1888 */ getValueTypeForKey(@onNull String name)1889 public final @Type int getValueTypeForKey(@NonNull String name) { 1890 Object value = mMap.get(name); 1891 if (value == null) { 1892 return TYPE_NULL; 1893 } else if (value instanceof Integer) { 1894 return TYPE_INTEGER; 1895 } else if (value instanceof Long) { 1896 return TYPE_LONG; 1897 } else if (value instanceof Float) { 1898 return TYPE_FLOAT; 1899 } else if (value instanceof String) { 1900 return TYPE_STRING; 1901 } else if (value instanceof ByteBuffer) { 1902 return TYPE_BYTE_BUFFER; 1903 } 1904 throw new RuntimeException("invalid value for key"); 1905 } 1906 1907 /** 1908 * A key prefix used together with a {@link MediaCodecInfo.CodecCapabilities} 1909 * feature name describing a required or optional feature for a codec capabilities 1910 * query. 1911 * The associated value is an integer, where non-0 value means the feature is 1912 * requested to be present, while 0 value means the feature is requested to be not 1913 * present. 1914 * @see MediaCodecList#findDecoderForFormat 1915 * @see MediaCodecList#findEncoderForFormat 1916 * @see MediaCodecInfo.CodecCapabilities#isFormatSupported 1917 * 1918 * @hide 1919 */ 1920 public static final String KEY_FEATURE_ = "feature-"; 1921 1922 /** 1923 * Returns the value of a numeric key. This is provided as a convenience method for keys 1924 * that may take multiple numeric types, such as {@link #KEY_FRAME_RATE}, or {@link 1925 * #KEY_I_FRAME_INTERVAL}. 1926 * 1927 * @return null if the key does not exist or the stored value for the key is null 1928 * @throws ClassCastException if the stored value for the key is ByteBuffer or String 1929 */ getNumber(@onNull String name)1930 public final @Nullable Number getNumber(@NonNull String name) { 1931 return (Number) mMap.get(name); 1932 } 1933 1934 /** 1935 * Returns the value of a numeric key, or the default value if the key is missing. 1936 * 1937 * @return defaultValue if the key does not exist or the stored value for the key is null 1938 * @throws ClassCastException if the stored value for the key is ByteBuffer or String 1939 */ getNumber(@onNull String name, @NonNull Number defaultValue)1940 public final @NonNull Number getNumber(@NonNull String name, @NonNull Number defaultValue) { 1941 Number ret = getNumber(name); 1942 return ret == null ? defaultValue : ret; 1943 } 1944 1945 /** 1946 * Returns the value of an integer key. 1947 * 1948 * @throws NullPointerException if the key does not exist or the stored value for the key is 1949 * null 1950 * @throws ClassCastException if the stored value for the key is long, float, ByteBuffer or 1951 * String 1952 */ getInteger(@onNull String name)1953 public final int getInteger(@NonNull String name) { 1954 return (int) mMap.get(name); 1955 } 1956 1957 /** 1958 * Returns the value of an integer key, or the default value if the key is missing. 1959 * 1960 * @return defaultValue if the key does not exist or the stored value for the key is null 1961 * @throws ClassCastException if the stored value for the key is long, float, ByteBuffer or 1962 * String 1963 */ getInteger(@onNull String name, int defaultValue)1964 public final int getInteger(@NonNull String name, int defaultValue) { 1965 try { 1966 return getInteger(name); 1967 } catch (NullPointerException e) { 1968 /* no such field or field is null */ 1969 return defaultValue; 1970 } 1971 } 1972 1973 /** 1974 * Returns the value of a long key. 1975 * 1976 * @throws NullPointerException if the key does not exist or the stored value for the key is 1977 * null 1978 * @throws ClassCastException if the stored value for the key is int, float, ByteBuffer or 1979 * String 1980 */ getLong(@onNull String name)1981 public final long getLong(@NonNull String name) { 1982 return (long) mMap.get(name); 1983 } 1984 1985 /** 1986 * Returns the value of a long key, or the default value if the key is missing. 1987 * 1988 * @return defaultValue if the key does not exist or the stored value for the key is null 1989 * @throws ClassCastException if the stored value for the key is int, float, ByteBuffer or 1990 * String 1991 */ getLong(@onNull String name, long defaultValue)1992 public final long getLong(@NonNull String name, long defaultValue) { 1993 try { 1994 return getLong(name); 1995 } catch (NullPointerException e) { 1996 /* no such field or field is null */ 1997 return defaultValue; 1998 } 1999 } 2000 2001 /** 2002 * Returns the value of a float key. 2003 * 2004 * @throws NullPointerException if the key does not exist or the stored value for the key is 2005 * null 2006 * @throws ClassCastException if the stored value for the key is int, long, ByteBuffer or 2007 * String 2008 */ getFloat(@onNull String name)2009 public final float getFloat(@NonNull String name) { 2010 return (float) mMap.get(name); 2011 } 2012 2013 /** 2014 * Returns the value of a float key, or the default value if the key is missing. 2015 * 2016 * @return defaultValue if the key does not exist or the stored value for the key is null 2017 * @throws ClassCastException if the stored value for the key is int, long, ByteBuffer or 2018 * String 2019 */ getFloat(@onNull String name, float defaultValue)2020 public final float getFloat(@NonNull String name, float defaultValue) { 2021 Object value = mMap.get(name); 2022 return value != null ? (float) value : defaultValue; 2023 } 2024 2025 /** 2026 * Returns the value of a string key. 2027 * 2028 * @return null if the key does not exist or the stored value for the key is null 2029 * @throws ClassCastException if the stored value for the key is int, long, float or ByteBuffer 2030 */ getString(@onNull String name)2031 public final @Nullable String getString(@NonNull String name) { 2032 return (String)mMap.get(name); 2033 } 2034 2035 /** 2036 * Returns the value of a string key, or the default value if the key is missing. 2037 * 2038 * @return defaultValue if the key does not exist or the stored value for the key is null 2039 * @throws ClassCastException if the stored value for the key is int, long, float or ByteBuffer 2040 */ getString(@onNull String name, @NonNull String defaultValue)2041 public final @NonNull String getString(@NonNull String name, @NonNull String defaultValue) { 2042 String ret = getString(name); 2043 return ret == null ? defaultValue : ret; 2044 } 2045 2046 /** 2047 * Returns the value of a ByteBuffer key. 2048 * 2049 * @return null if the key does not exist or the stored value for the key is null 2050 * @throws ClassCastException if the stored value for the key is int, long, float or String 2051 */ getByteBuffer(@onNull String name)2052 public final @Nullable ByteBuffer getByteBuffer(@NonNull String name) { 2053 return (ByteBuffer)mMap.get(name); 2054 } 2055 2056 /** 2057 * Returns the value of a ByteBuffer key, or the default value if the key is missing. 2058 * 2059 * @return defaultValue if the key does not exist or the stored value for the key is null 2060 * @throws ClassCastException if the stored value for the key is int, long, float or String 2061 */ getByteBuffer( @onNull String name, @NonNull ByteBuffer defaultValue)2062 public final @NonNull ByteBuffer getByteBuffer( 2063 @NonNull String name, @NonNull ByteBuffer defaultValue) { 2064 ByteBuffer ret = getByteBuffer(name); 2065 return ret == null ? defaultValue : ret; 2066 } 2067 2068 /** 2069 * Returns whether a feature is to be enabled ({@code true}) or disabled 2070 * ({@code false}). 2071 * 2072 * @param feature the name of a {@link MediaCodecInfo.CodecCapabilities} feature. 2073 * 2074 * @throws IllegalArgumentException if the feature was neither set to be enabled 2075 * nor to be disabled. 2076 */ getFeatureEnabled(@onNull String feature)2077 public boolean getFeatureEnabled(@NonNull String feature) { 2078 Integer enabled = (Integer)mMap.get(KEY_FEATURE_ + feature); 2079 if (enabled == null) { 2080 throw new IllegalArgumentException("feature is not specified"); 2081 } 2082 return enabled != 0; 2083 } 2084 2085 /** 2086 * Sets the value of an integer key. 2087 */ setInteger(@onNull String name, int value)2088 public final void setInteger(@NonNull String name, int value) { 2089 mMap.put(name, value); 2090 } 2091 2092 /** 2093 * Sets the value of a long key. 2094 */ setLong(@onNull String name, long value)2095 public final void setLong(@NonNull String name, long value) { 2096 mMap.put(name, value); 2097 } 2098 2099 /** 2100 * Sets the value of a float key. 2101 */ setFloat(@onNull String name, float value)2102 public final void setFloat(@NonNull String name, float value) { 2103 mMap.put(name, value); 2104 } 2105 2106 /** 2107 * Sets the value of a string key. 2108 * <p> 2109 * If value is {@code null}, it sets a null value that behaves similarly to a missing key. 2110 * This could be used prior to API level {@link android os.Build.VERSION_CODES#Q} to effectively 2111 * remove a key. 2112 */ setString(@onNull String name, @Nullable String value)2113 public final void setString(@NonNull String name, @Nullable String value) { 2114 mMap.put(name, value); 2115 } 2116 2117 /** 2118 * Sets the value of a ByteBuffer key. 2119 * <p> 2120 * If value is {@code null}, it sets a null value that behaves similarly to a missing key. 2121 * This could be used prior to API level {@link android os.Build.VERSION_CODES#Q} to effectively 2122 * remove a key. 2123 */ setByteBuffer(@onNull String name, @Nullable ByteBuffer bytes)2124 public final void setByteBuffer(@NonNull String name, @Nullable ByteBuffer bytes) { 2125 mMap.put(name, bytes); 2126 } 2127 2128 /** 2129 * Removes a value of a given key if present. Has no effect if the key is not present. 2130 */ removeKey(@onNull String name)2131 public final void removeKey(@NonNull String name) { 2132 // exclude feature mappings 2133 if (!name.startsWith(KEY_FEATURE_)) { 2134 mMap.remove(name); 2135 } 2136 } 2137 2138 /** 2139 * Removes a given feature setting if present. Has no effect if the feature setting is not 2140 * present. 2141 */ removeFeature(@onNull String name)2142 public final void removeFeature(@NonNull String name) { 2143 mMap.remove(KEY_FEATURE_ + name); 2144 } 2145 2146 /** 2147 * A Partial set view for a portion of the keys in a MediaFormat object. 2148 * 2149 * This class is needed as we want to return a portion of the actual format keys in getKeys() 2150 * and another portion of the keys in getFeatures(), and still allow the view properties. 2151 */ 2152 private abstract class FilteredMappedKeySet extends AbstractSet<String> { 2153 private Set<String> mKeys; 2154 2155 // Returns true if this set should include this key keepKey(String key)2156 abstract protected boolean keepKey(String key); 2157 2158 // Maps a key from the underlying key set into its new value in this key set mapKeyToItem(String key)2159 abstract protected String mapKeyToItem(String key); 2160 2161 // Maps a key from this key set into its original value in the underlying key set mapItemToKey(String item)2162 abstract protected String mapItemToKey(String item); 2163 FilteredMappedKeySet()2164 public FilteredMappedKeySet() { 2165 mKeys = mMap.keySet(); 2166 } 2167 2168 // speed up contains and remove from abstract implementation (that would iterate 2169 // over each element) 2170 @Override contains(Object o)2171 public boolean contains(Object o) { 2172 if (o instanceof String) { 2173 String key = mapItemToKey((String)o); 2174 return keepKey(key) && mKeys.contains(key); 2175 } 2176 return false; 2177 } 2178 2179 @Override remove(Object o)2180 public boolean remove(Object o) { 2181 if (o instanceof String) { 2182 String key = mapItemToKey((String)o); 2183 if (keepKey(key) && mKeys.remove(key)) { 2184 mMap.remove(key); 2185 return true; 2186 } 2187 } 2188 return false; 2189 } 2190 2191 private class KeyIterator implements Iterator<String> { 2192 Iterator<String> mIterator; 2193 String mLast; 2194 KeyIterator()2195 public KeyIterator() { 2196 // We must create a copy of the filtered stream, as remove operation has to modify 2197 // the underlying data structure (mMap), so the iterator's operation is undefined. 2198 // Use a list as it is likely less memory consuming than the other alternative: set. 2199 mIterator = 2200 mKeys.stream().filter(k -> keepKey(k)).collect(Collectors.toList()).iterator(); 2201 } 2202 2203 @Override hasNext()2204 public boolean hasNext() { 2205 return mIterator.hasNext(); 2206 } 2207 2208 @Override next()2209 public String next() { 2210 mLast = mIterator.next(); 2211 return mapKeyToItem(mLast); 2212 } 2213 2214 @Override remove()2215 public void remove() { 2216 mIterator.remove(); 2217 mMap.remove(mLast); 2218 } 2219 } 2220 2221 @Override iterator()2222 public Iterator<String> iterator() { 2223 return new KeyIterator(); 2224 } 2225 2226 @Override size()2227 public int size() { 2228 return (int) mKeys.stream().filter(this::keepKey).count(); 2229 } 2230 } 2231 2232 /** 2233 * A Partial set view for a portion of the keys in a MediaFormat object for keys that 2234 * don't start with a prefix, such as "feature-" 2235 */ 2236 private class UnprefixedKeySet extends FilteredMappedKeySet { 2237 private String mPrefix; 2238 UnprefixedKeySet(String prefix)2239 public UnprefixedKeySet(String prefix) { 2240 super(); 2241 mPrefix = prefix; 2242 } 2243 keepKey(String key)2244 protected boolean keepKey(String key) { 2245 return !key.startsWith(mPrefix); 2246 } 2247 mapKeyToItem(String key)2248 protected String mapKeyToItem(String key) { 2249 return key; 2250 } 2251 mapItemToKey(String item)2252 protected String mapItemToKey(String item) { 2253 return item; 2254 } 2255 } 2256 2257 /** 2258 * A Partial set view for a portion of the keys in a MediaFormat object for keys that 2259 * start with a prefix, such as "feature-", with the prefix removed 2260 */ 2261 private class PrefixedKeySetWithPrefixRemoved extends FilteredMappedKeySet { 2262 private String mPrefix; 2263 private int mPrefixLength; 2264 PrefixedKeySetWithPrefixRemoved(String prefix)2265 public PrefixedKeySetWithPrefixRemoved(String prefix) { 2266 super(); 2267 mPrefix = prefix; 2268 mPrefixLength = prefix.length(); 2269 } 2270 keepKey(String key)2271 protected boolean keepKey(String key) { 2272 return key.startsWith(mPrefix); 2273 } 2274 mapKeyToItem(String key)2275 protected String mapKeyToItem(String key) { 2276 return key.substring(mPrefixLength); 2277 } 2278 mapItemToKey(String item)2279 protected String mapItemToKey(String item) { 2280 return mPrefix + item; 2281 } 2282 } 2283 2284 2285 /** 2286 * Returns a {@link java.util.Set Set} view of the keys contained in this MediaFormat. 2287 * 2288 * The set is backed by the MediaFormat object, so changes to the format are reflected in the 2289 * set, and vice-versa. If the format is modified while an iteration over the set is in progress 2290 * (except through the iterator's own remove operation), the results of the iteration are 2291 * undefined. The set supports element removal, which removes the corresponding mapping from the 2292 * format, via the Iterator.remove, Set.remove, removeAll, retainAll, and clear operations. 2293 * It does not support the add or addAll operations. 2294 */ getKeys()2295 public final @NonNull java.util.Set<String> getKeys() { 2296 return new UnprefixedKeySet(KEY_FEATURE_); 2297 } 2298 2299 /** 2300 * Returns a {@link java.util.Set Set} view of the features contained in this MediaFormat. 2301 * 2302 * The set is backed by the MediaFormat object, so changes to the format are reflected in the 2303 * set, and vice-versa. If the format is modified while an iteration over the set is in progress 2304 * (except through the iterator's own remove operation), the results of the iteration are 2305 * undefined. The set supports element removal, which removes the corresponding mapping from the 2306 * format, via the Iterator.remove, Set.remove, removeAll, retainAll, and clear operations. 2307 * It does not support the add or addAll operations. 2308 */ getFeatures()2309 public final @NonNull java.util.Set<String> getFeatures() { 2310 return new PrefixedKeySetWithPrefixRemoved(KEY_FEATURE_); 2311 } 2312 2313 /** 2314 * Create a copy of a media format object. 2315 */ MediaFormat(@onNull MediaFormat other)2316 public MediaFormat(@NonNull MediaFormat other) { 2317 this(); 2318 mMap.putAll(other.mMap); 2319 } 2320 2321 /** 2322 * Sets whether a feature is to be enabled ({@code true}) or disabled 2323 * ({@code false}). 2324 * 2325 * If {@code enabled} is {@code true}, the feature is requested to be present. 2326 * Otherwise, the feature is requested to be not present. 2327 * 2328 * @param feature the name of a {@link MediaCodecInfo.CodecCapabilities} feature. 2329 * 2330 * @see MediaCodecList#findDecoderForFormat 2331 * @see MediaCodecList#findEncoderForFormat 2332 * @see MediaCodecInfo.CodecCapabilities#isFormatSupported 2333 */ setFeatureEnabled(@onNull String feature, boolean enabled)2334 public void setFeatureEnabled(@NonNull String feature, boolean enabled) { 2335 setInteger(KEY_FEATURE_ + feature, enabled ? 1 : 0); 2336 } 2337 2338 /** 2339 * Creates a minimal audio format. 2340 * @param mime The mime type of the content. 2341 * @param sampleRate The sampling rate of the content. 2342 * @param channelCount The number of audio channels in the content. 2343 */ createAudioFormat( @onNull String mime, int sampleRate, int channelCount)2344 public static final @NonNull MediaFormat createAudioFormat( 2345 @NonNull String mime, 2346 int sampleRate, 2347 int channelCount) { 2348 MediaFormat format = new MediaFormat(); 2349 format.setString(KEY_MIME, mime); 2350 format.setInteger(KEY_SAMPLE_RATE, sampleRate); 2351 format.setInteger(KEY_CHANNEL_COUNT, channelCount); 2352 2353 return format; 2354 } 2355 2356 /** 2357 * Creates a minimal subtitle format. 2358 * @param mime The mime type of the content. 2359 * @param language The language of the content, using either ISO 639-1 or 639-2/T 2360 * codes. Specify null or "und" if language information is only included 2361 * in the content. (This will also work if there are multiple language 2362 * tracks in the content.) 2363 */ createSubtitleFormat( @onNull String mime, String language)2364 public static final @NonNull MediaFormat createSubtitleFormat( 2365 @NonNull String mime, 2366 String language) { 2367 MediaFormat format = new MediaFormat(); 2368 format.setString(KEY_MIME, mime); 2369 format.setString(KEY_LANGUAGE, language); 2370 2371 return format; 2372 } 2373 2374 /** 2375 * Creates a minimal video format. 2376 * @param mime The mime type of the content. 2377 * @param width The width of the content (in pixels) 2378 * @param height The height of the content (in pixels) 2379 */ createVideoFormat( @onNull String mime, int width, int height)2380 public static final @NonNull MediaFormat createVideoFormat( 2381 @NonNull String mime, 2382 int width, 2383 int height) { 2384 MediaFormat format = new MediaFormat(); 2385 format.setString(KEY_MIME, mime); 2386 format.setInteger(KEY_WIDTH, width); 2387 format.setInteger(KEY_HEIGHT, height); 2388 2389 return format; 2390 } 2391 2392 @Override toString()2393 public @NonNull String toString() { 2394 return mMap.toString(); 2395 } 2396 } 2397
