server/helpers/ffmpeg-utils.ts

   1 import * as ffmpeg from 'fluent-ffmpeg'
   2 import { dirname, join } from 'path'
   3 import { getTargetBitrate, getMaxBitrate, VideoResolution } from '../../shared/models/videos'
   4 import { FFMPEG_NICE, VIDEO_TRANSCODING_FPS } from '../initializers/constants'
   5 import { processImage } from './image-utils'
   6 import { logger } from './logger'
   7 import { checkFFmpegEncoders } from '../initializers/checker-before-init'
   8 import { readFile, remove, writeFile } from 'fs-extra'
   9 import { CONFIG } from '../initializers/config'
  10
  11 function computeResolutionsToTranscode (videoFileHeight: number) {
  12   const resolutionsEnabled: number[] = []
  13   const configResolutions = CONFIG.TRANSCODING.RESOLUTIONS
  14
  15   // Put in the order we want to proceed jobs
  16   const resolutions = [
  17     VideoResolution.H_NOVIDEO,
  18     VideoResolution.H_480P,
  19     VideoResolution.H_360P,
  20     VideoResolution.H_720P,
  21     VideoResolution.H_240P,
  22     VideoResolution.H_1080P,
  23     VideoResolution.H_4K
  24   ]
  25
  26   for (const resolution of resolutions) {
  27     if (configResolutions[ resolution + 'p' ] === true && videoFileHeight > resolution) {
  28       resolutionsEnabled.push(resolution)
  29     }
  30   }
  31
  32   return resolutionsEnabled
  33 }
  34
  35 async function getVideoStreamSize (path: string) {
  36   const videoStream = await getVideoStreamFromFile(path)
  37
  38   return videoStream === null
  39     ? { width: 0, height: 0 }
  40     : { width: videoStream.width, height: videoStream.height }
  41 }
  42
  43 async function getVideoStreamCodec (path: string) {
  44   const videoStream = await getVideoStreamFromFile(path)
  45
  46   if (!videoStream) return ''
  47
  48   const videoCodec = videoStream.codec_tag_string
  49
  50   const baseProfileMatrix = {
  51     'High': '6400',
  52     'Main': '4D40',
  53     'Baseline': '42E0'
  54   }
  55
  56   let baseProfile = baseProfileMatrix[videoStream.profile]
  57   if (!baseProfile) {
  58     logger.warn('Cannot get video profile codec of %s.', path, { videoStream })
  59     baseProfile = baseProfileMatrix['High'] // Fallback
  60   }
  61
  62   const level = videoStream.level.toString(16)
  63
  64   return `${videoCodec}.${baseProfile}${level}`
  65 }
  66
  67 async function getAudioStreamCodec (path: string) {
  68   const { audioStream } = await audio.get(path)
  69
  70   if (!audioStream) return ''
  71
  72   const audioCodec = audioStream.codec_name
  73   if (audioCodec === 'aac') return 'mp4a.40.2'
  74
  75   logger.warn('Cannot get audio codec of %s.', path, { audioStream })
  76
  77   return 'mp4a.40.2' // Fallback
  78 }
  79
  80 async function getVideoFileResolution (path: string) {
  81   const size = await getVideoStreamSize(path)
  82
  83   return {
  84     videoFileResolution: Math.min(size.height, size.width),
  85     isPortraitMode: size.height > size.width
  86   }
  87 }
  88
  89 async function getVideoFileFPS (path: string) {
  90   const videoStream = await getVideoStreamFromFile(path)
  91   if (videoStream === null) return 0
  92
  93   for (const key of [ 'avg_frame_rate', 'r_frame_rate' ]) {
  94     const valuesText: string = videoStream[ key ]
  95     if (!valuesText) continue
  96
  97     const [ frames, seconds ] = valuesText.split('/')
  98     if (!frames || !seconds) continue
  99
 100     const result = parseInt(frames, 10) / parseInt(seconds, 10)
 101     if (result > 0) return Math.round(result)
 102   }
 103
 104   return 0
 105 }
 106
 107 async function getVideoFileBitrate (path: string) {
 108   return new Promise<number>((res, rej) => {
 109     ffmpeg.ffprobe(path, (err, metadata) => {
 110       if (err) return rej(err)
 111
 112       return res(metadata.format.bit_rate)
 113     })
 114   })
 115 }
 116
 117 function getDurationFromVideoFile (path: string) {
 118   return new Promise<number>((res, rej) => {
 119     ffmpeg.ffprobe(path, (err, metadata) => {
 120       if (err) return rej(err)
 121
 122       return res(Math.floor(metadata.format.duration))
 123     })
 124   })
 125 }
 126
 127 async function generateImageFromVideoFile (fromPath: string, folder: string, imageName: string, size: { width: number, height: number }) {
 128   const pendingImageName = 'pending-' + imageName
 129
 130   const options = {
 131     filename: pendingImageName,
 132     count: 1,
 133     folder
 134   }
 135
 136   const pendingImagePath = join(folder, pendingImageName)
 137
 138   try {
 139     await new Promise<string>((res, rej) => {
 140       ffmpeg(fromPath, { niceness: FFMPEG_NICE.THUMBNAIL })
 141         .on('error', rej)
 142         .on('end', () => res(imageName))
 143         .thumbnail(options)
 144     })
 145
 146     const destination = join(folder, imageName)
 147     await processImage(pendingImagePath, destination, size)
 148   } catch (err) {
 149     logger.error('Cannot generate image from video %s.', fromPath, { err })
 150
 151     try {
 152       await remove(pendingImagePath)
 153     } catch (err) {
 154       logger.debug('Cannot remove pending image path after generation error.', { err })
 155     }
 156   }
 157 }
 158
 159 type TranscodeOptionsType = 'hls' | 'quick-transcode' | 'video' | 'merge-audio' | 'only-audio'
 160
 161 interface BaseTranscodeOptions {
 162   type: TranscodeOptionsType
 163   inputPath: string
 164   outputPath: string
 165   resolution: VideoResolution
 166   isPortraitMode?: boolean
 167 }
 168
 169 interface HLSTranscodeOptions extends BaseTranscodeOptions {
 170   type: 'hls'
 171   copyCodecs: boolean
 172   hlsPlaylist: {
 173     videoFilename: string
 174   }
 175 }
 176
 177 interface QuickTranscodeOptions extends BaseTranscodeOptions {
 178   type: 'quick-transcode'
 179 }
 180
 181 interface VideoTranscodeOptions extends BaseTranscodeOptions {
 182   type: 'video'
 183 }
 184
 185 interface MergeAudioTranscodeOptions extends BaseTranscodeOptions {
 186   type: 'merge-audio'
 187   audioPath: string
 188 }
 189
 190 interface OnlyAudioTranscodeOptions extends BaseTranscodeOptions {
 191   type: 'only-audio'
 192 }
 193
 194 type TranscodeOptions = HLSTranscodeOptions
 195   | VideoTranscodeOptions
 196   | MergeAudioTranscodeOptions
 197   | OnlyAudioTranscodeOptions
 198   | QuickTranscodeOptions
 199
 200 function transcode (options: TranscodeOptions) {
 201   return new Promise<void>(async (res, rej) => {
 202     try {
 203       let command = ffmpeg(options.inputPath, { niceness: FFMPEG_NICE.TRANSCODING })
 204         .output(options.outputPath)
 205
 206       if (options.type === 'quick-transcode') {
 207         command = await buildQuickTranscodeCommand(command)
 208       } else if (options.type === 'hls') {
 209         command = await buildHLSCommand(command, options)
 210       } else if (options.type === 'merge-audio') {
 211         command = await buildAudioMergeCommand(command, options)
 212       } else if (options.type === 'only-audio') {
 213         command = await buildOnlyAudioCommand(command, options)
 214       } else {
 215         command = await buildx264Command(command, options)
 216       }
 217
 218       if (CONFIG.TRANSCODING.THREADS > 0) {
 219         // if we don't set any threads ffmpeg will chose automatically
 220         command = command.outputOption('-threads ' + CONFIG.TRANSCODING.THREADS)
 221       }
 222
 223       command
 224         .on('error', (err, stdout, stderr) => {
 225           logger.error('Error in transcoding job.', { stdout, stderr })
 226           return rej(err)
 227         })
 228         .on('end', () => {
 229           return fixHLSPlaylistIfNeeded(options)
 230             .then(() => res())
 231             .catch(err => rej(err))
 232         })
 233         .run()
 234     } catch (err) {
 235       return rej(err)
 236     }
 237   })
 238 }
 239
 240 async function canDoQuickTranscode (path: string): Promise<boolean> {
 241   // NOTE: This could be optimized by running ffprobe only once (but it runs fast anyway)
 242   const videoStream = await getVideoStreamFromFile(path)
 243   const parsedAudio = await audio.get(path)
 244   const fps = await getVideoFileFPS(path)
 245   const bitRate = await getVideoFileBitrate(path)
 246   const resolution = await getVideoFileResolution(path)
 247
 248   // check video params
 249   if (videoStream == null) return false
 250   if (videoStream[ 'codec_name' ] !== 'h264') return false
 251   if (videoStream[ 'pix_fmt' ] !== 'yuv420p') return false
 252   if (fps < VIDEO_TRANSCODING_FPS.MIN || fps > VIDEO_TRANSCODING_FPS.MAX) return false
 253   if (bitRate > getMaxBitrate(resolution.videoFileResolution, fps, VIDEO_TRANSCODING_FPS)) return false
 254
 255   // check audio params (if audio stream exists)
 256   if (parsedAudio.audioStream) {
 257     if (parsedAudio.audioStream[ 'codec_name' ] !== 'aac') return false
 258
 259     const maxAudioBitrate = audio.bitrate[ 'aac' ](parsedAudio.audioStream[ 'bit_rate' ])
 260     if (maxAudioBitrate !== -1 && parsedAudio.audioStream[ 'bit_rate' ] > maxAudioBitrate) return false
 261   }
 262
 263   return true
 264 }
 265
 266 // ---------------------------------------------------------------------------
 267
 268 export {
 269   getVideoStreamCodec,
 270   getAudioStreamCodec,
 271   getVideoStreamSize,
 272   getVideoFileResolution,
 273   getDurationFromVideoFile,
 274   generateImageFromVideoFile,
 275   TranscodeOptions,
 276   TranscodeOptionsType,
 277   transcode,
 278   getVideoFileFPS,
 279   computeResolutionsToTranscode,
 280   audio,
 281   getVideoFileBitrate,
 282   canDoQuickTranscode
 283 }
 284
 285 // ---------------------------------------------------------------------------
 286
 287 async function buildx264Command (command: ffmpeg.FfmpegCommand, options: TranscodeOptions) {
 288   let fps = await getVideoFileFPS(options.inputPath)
 289   if (
 290     // On small/medium resolutions, limit FPS
 291     options.resolution !== undefined &&
 292     options.resolution < VIDEO_TRANSCODING_FPS.KEEP_ORIGIN_FPS_RESOLUTION_MIN &&
 293     fps > VIDEO_TRANSCODING_FPS.AVERAGE ||
 294     // If the video is doesn't match had standard
 295     !VIDEO_TRANSCODING_FPS.HD_STANDARD.map(value => fps % value).includes(0)
 296   ) {
 297     // Get closest standard framerate by modulo: downsampling has to be done to a divisor of the nominal fps value
 298     fps = VIDEO_TRANSCODING_FPS.STANDARD.sort((a, b) => fps % a - fps % b)[0]
 299   }
 300
 301   command = await presetH264(command, options.inputPath, options.resolution, fps)
 302
 303   if (options.resolution !== undefined) {
 304     // '?x720' or '720x?' for example
 305     const size = options.isPortraitMode === true ? `${options.resolution}x?` : `?x${options.resolution}`
 306     command = command.size(size)
 307   }
 308
 309   if (fps) {
 310     // Hard FPS limits
 311     if (fps > VIDEO_TRANSCODING_FPS.MAX) fps = VIDEO_TRANSCODING_FPS.HD_STANDARD.sort((a, b) => fps % a - fps % b)[0]
 312     else if (fps < VIDEO_TRANSCODING_FPS.MIN) fps = VIDEO_TRANSCODING_FPS.MIN
 313
 314     command = command.withFPS(fps)
 315   }
 316
 317   return command
 318 }
 319
 320 async function buildAudioMergeCommand (command: ffmpeg.FfmpegCommand, options: MergeAudioTranscodeOptions) {
 321   command = command.loop(undefined)
 322
 323   command = await presetH264VeryFast(command, options.audioPath, options.resolution)
 324
 325   command = command.input(options.audioPath)
 326                    .videoFilter('scale=trunc(iw/2)*2:trunc(ih/2)*2') // Avoid "height not divisible by 2" error
 327                    .outputOption('-tune stillimage')
 328                    .outputOption('-shortest')
 329
 330   return command
 331 }
 332
 333 async function buildOnlyAudioCommand (command: ffmpeg.FfmpegCommand, options: OnlyAudioTranscodeOptions) {
 334   command = await presetOnlyAudio(command)
 335
 336   return command
 337 }
 338
 339 async function buildQuickTranscodeCommand (command: ffmpeg.FfmpegCommand) {
 340   command = await presetCopy(command)
 341
 342   command = command.outputOption('-map_metadata -1') // strip all metadata
 343                    .outputOption('-movflags faststart')
 344
 345   return command
 346 }
 347
 348 async function buildHLSCommand (command: ffmpeg.FfmpegCommand, options: HLSTranscodeOptions) {
 349   const videoPath = getHLSVideoPath(options)
 350
 351   if (options.copyCodecs) command = await presetCopy(command)
 352   else command = await buildx264Command(command, options)
 353
 354   command = command.outputOption('-hls_time 4')
 355                    .outputOption('-hls_list_size 0')
 356                    .outputOption('-hls_playlist_type vod')
 357                    .outputOption('-hls_segment_filename ' + videoPath)
 358                    .outputOption('-hls_segment_type fmp4')
 359                    .outputOption('-f hls')
 360                    .outputOption('-hls_flags single_file')
 361
 362   return command
 363 }
 364
 365 function getHLSVideoPath (options: HLSTranscodeOptions) {
 366   return `${dirname(options.outputPath)}/${options.hlsPlaylist.videoFilename}`
 367 }
 368
 369 async function fixHLSPlaylistIfNeeded (options: TranscodeOptions) {
 370   if (options.type !== 'hls') return
 371
 372   const fileContent = await readFile(options.outputPath)
 373
 374   const videoFileName = options.hlsPlaylist.videoFilename
 375   const videoFilePath = getHLSVideoPath(options)
 376
 377   // Fix wrong mapping with some ffmpeg versions
 378   const newContent = fileContent.toString()
 379                                 .replace(`#EXT-X-MAP:URI="${videoFilePath}",`, `#EXT-X-MAP:URI="${videoFileName}",`)
 380
 381   await writeFile(options.outputPath, newContent)
 382 }
 383
 384 function getVideoStreamFromFile (path: string) {
 385   return new Promise<any>((res, rej) => {
 386     ffmpeg.ffprobe(path, (err, metadata) => {
 387       if (err) return rej(err)
 388
 389       const videoStream = metadata.streams.find(s => s.codec_type === 'video')
 390       return res(videoStream || null)
 391     })
 392   })
 393 }
 394
 395 /**
 396  * A slightly customised version of the 'veryfast' x264 preset
 397  *
 398  * The veryfast preset is right in the sweet spot of performance
 399  * and quality. Superfast and ultrafast will give you better
 400  * performance, but then quality is noticeably worse.
 401  */
 402 async function presetH264VeryFast (command: ffmpeg.FfmpegCommand, input: string, resolution: VideoResolution, fps?: number) {
 403   let localCommand = await presetH264(command, input, resolution, fps)
 404
 405   localCommand = localCommand.outputOption('-preset:v veryfast')
 406
 407   /*
 408   MAIN reference: https://slhck.info/video/2017/03/01/rate-control.html
 409   Our target situation is closer to a livestream than a stream,
 410   since we want to reduce as much a possible the encoding burden,
 411   although not to the point of a livestream where there is a hard
 412   constraint on the frames per second to be encoded.
 413   */
 414
 415   return localCommand
 416 }
 417
 418 /**
 419  * A toolbox to play with audio
 420  */
 421 namespace audio {
 422   export const get = (videoPath: string) => {
 423     // without position, ffprobe considers the last input only
 424     // we make it consider the first input only
 425     // if you pass a file path to pos, then ffprobe acts on that file directly
 426     return new Promise<{ absolutePath: string, audioStream?: any }>((res, rej) => {
 427
 428       function parseFfprobe (err: any, data: ffmpeg.FfprobeData) {
 429         if (err) return rej(err)
 430
 431         if ('streams' in data) {
 432           const audioStream = data.streams.find(stream => stream[ 'codec_type' ] === 'audio')
 433           if (audioStream) {
 434             return res({
 435               absolutePath: data.format.filename,
 436               audioStream
 437             })
 438           }
 439         }
 440
 441         return res({ absolutePath: data.format.filename })
 442       }
 443
 444       return ffmpeg.ffprobe(videoPath, parseFfprobe)
 445     })
 446   }
 447
 448   export namespace bitrate {
 449     const baseKbitrate = 384
 450
 451     const toBits = (kbits: number) => kbits * 8000
 452
 453     export const aac = (bitrate: number): number => {
 454       switch (true) {
 455         case bitrate > toBits(baseKbitrate):
 456           return baseKbitrate
 457
 458         default:
 459           return -1 // we interpret it as a signal to copy the audio stream as is
 460       }
 461     }
 462
 463     export const mp3 = (bitrate: number): number => {
 464       /*
 465       a 192kbit/sec mp3 doesn't hold as much information as a 192kbit/sec aac.
 466       That's why, when using aac, we can go to lower kbit/sec. The equivalences
 467       made here are not made to be accurate, especially with good mp3 encoders.
 468       */
 469       switch (true) {
 470         case bitrate <= toBits(192):
 471           return 128
 472
 473         case bitrate <= toBits(384):
 474           return 256
 475
 476         default:
 477           return baseKbitrate
 478       }
 479     }
 480   }
 481 }
 482
 483 /**
 484  * Standard profile, with variable bitrate audio and faststart.
 485  *
 486  * As for the audio, quality '5' is the highest and ensures 96-112kbps/channel
 487  * See https://trac.ffmpeg.org/wiki/Encode/AAC#fdk_vbr
 488  */
 489 async function presetH264 (command: ffmpeg.FfmpegCommand, input: string, resolution: VideoResolution, fps?: number) {
 490   let localCommand = command
 491     .format('mp4')
 492     .videoCodec('libx264')
 493     .outputOption('-level 3.1') // 3.1 is the minimal resource allocation for our highest supported resolution
 494     .outputOption('-b_strategy 1') // NOTE: b-strategy 1 - heuristic algorithm, 16 is optimal B-frames for it
 495     .outputOption('-bf 16') // NOTE: Why 16: https://github.com/Chocobozzz/PeerTube/pull/774. b-strategy 2 -> B-frames<16
 496     .outputOption('-pix_fmt yuv420p') // allows import of source material with incompatible pixel formats (e.g. MJPEG video)
 497     .outputOption('-map_metadata -1') // strip all metadata
 498     .outputOption('-movflags faststart')
 499
 500   const parsedAudio = await audio.get(input)
 501
 502   if (!parsedAudio.audioStream) {
 503     localCommand = localCommand.noAudio()
 504   } else if ((await checkFFmpegEncoders()).get('libfdk_aac')) { // we favor VBR, if a good AAC encoder is available
 505     localCommand = localCommand
 506       .audioCodec('libfdk_aac')
 507       .audioQuality(5)
 508   } else {
 509     // we try to reduce the ceiling bitrate by making rough matches of bitrates
 510     // of course this is far from perfect, but it might save some space in the end
 511     localCommand = localCommand.audioCodec('aac')
 512
 513     const audioCodecName = parsedAudio.audioStream[ 'codec_name' ]
 514
 515     if (audio.bitrate[ audioCodecName ]) {
 516       const bitrate = audio.bitrate[ audioCodecName ](parsedAudio.audioStream[ 'bit_rate' ])
 517       if (bitrate !== undefined && bitrate !== -1) localCommand = localCommand.audioBitrate(bitrate)
 518     }
 519   }
 520
 521   if (fps) {
 522     // Constrained Encoding (VBV)
 523     // https://slhck.info/video/2017/03/01/rate-control.html
 524     // https://trac.ffmpeg.org/wiki/Limiting%20the%20output%20bitrate
 525     const targetBitrate = getTargetBitrate(resolution, fps, VIDEO_TRANSCODING_FPS)
 526     localCommand = localCommand.outputOptions([ `-maxrate ${targetBitrate}`, `-bufsize ${targetBitrate * 2}` ])
 527
 528     // Keyframe interval of 2 seconds for faster seeking and resolution switching.
 529     // https://streaminglearningcenter.com/blogs/whats-the-right-keyframe-interval.html
 530     // https://superuser.com/a/908325
 531     localCommand = localCommand.outputOption(`-g ${fps * 2}`)
 532   }
 533
 534   return localCommand
 535 }
 536
 537 async function presetCopy (command: ffmpeg.FfmpegCommand): Promise<ffmpeg.FfmpegCommand> {
 538   return command
 539     .format('mp4')
 540     .videoCodec('copy')
 541     .audioCodec('copy')
 542 }
 543
 544 async function presetOnlyAudio (command: ffmpeg.FfmpegCommand): Promise<ffmpeg.FfmpegCommand> {
 545   return command
 546     .format('mp4')
 547     .audioCodec('copy')
 548     .noVideo()
 549 }