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