]> git.immae.eu Git - github/Chocobozzz/PeerTube.git/blob - server/helpers/ffmpeg-utils.ts
Don't quick transcode with the wrong pixel format
[github/Chocobozzz/PeerTube.git] / 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_480P,
18 VideoResolution.H_360P,
19 VideoResolution.H_720P,
20 VideoResolution.H_240P,
21 VideoResolution.H_1080P,
22 VideoResolution.H_4K
23 ]
24
25 for (const resolution of resolutions) {
26 if (configResolutions[ resolution + 'p' ] === true && videoFileHeight > resolution) {
27 resolutionsEnabled.push(resolution)
28 }
29 }
30
31 return resolutionsEnabled
32 }
33
34 async function getVideoFileSize (path: string) {
35 const videoStream = await getVideoStreamFromFile(path)
36
37 return {
38 width: videoStream.width,
39 height: videoStream.height
40 }
41 }
42
43 async function getVideoFileResolution (path: string) {
44 const size = await getVideoFileSize(path)
45
46 return {
47 videoFileResolution: Math.min(size.height, size.width),
48 isPortraitMode: size.height > size.width
49 }
50 }
51
52 async function getVideoFileFPS (path: string) {
53 const videoStream = await getVideoStreamFromFile(path)
54
55 for (const key of [ 'avg_frame_rate', 'r_frame_rate' ]) {
56 const valuesText: string = videoStream[key]
57 if (!valuesText) continue
58
59 const [ frames, seconds ] = valuesText.split('/')
60 if (!frames || !seconds) continue
61
62 const result = parseInt(frames, 10) / parseInt(seconds, 10)
63 if (result > 0) return Math.round(result)
64 }
65
66 return 0
67 }
68
69 async function getVideoFileBitrate (path: string) {
70 return new Promise<number>((res, rej) => {
71 ffmpeg.ffprobe(path, (err, metadata) => {
72 if (err) return rej(err)
73
74 return res(metadata.format.bit_rate)
75 })
76 })
77 }
78
79 function getDurationFromVideoFile (path: string) {
80 return new Promise<number>((res, rej) => {
81 ffmpeg.ffprobe(path, (err, metadata) => {
82 if (err) return rej(err)
83
84 return res(Math.floor(metadata.format.duration))
85 })
86 })
87 }
88
89 async function generateImageFromVideoFile (fromPath: string, folder: string, imageName: string, size: { width: number, height: number }) {
90 const pendingImageName = 'pending-' + imageName
91
92 const options = {
93 filename: pendingImageName,
94 count: 1,
95 folder
96 }
97
98 const pendingImagePath = join(folder, pendingImageName)
99
100 try {
101 await new Promise<string>((res, rej) => {
102 ffmpeg(fromPath, { niceness: FFMPEG_NICE.THUMBNAIL })
103 .on('error', rej)
104 .on('end', () => res(imageName))
105 .thumbnail(options)
106 })
107
108 const destination = join(folder, imageName)
109 await processImage(pendingImagePath, destination, size)
110 } catch (err) {
111 logger.error('Cannot generate image from video %s.', fromPath, { err })
112
113 try {
114 await remove(pendingImagePath)
115 } catch (err) {
116 logger.debug('Cannot remove pending image path after generation error.', { err })
117 }
118 }
119 }
120
121 type TranscodeOptionsType = 'hls' | 'quick-transcode' | 'video' | 'merge-audio'
122
123 interface BaseTranscodeOptions {
124 type: TranscodeOptionsType
125 inputPath: string
126 outputPath: string
127 resolution: VideoResolution
128 isPortraitMode?: boolean
129 }
130
131 interface HLSTranscodeOptions extends BaseTranscodeOptions {
132 type: 'hls'
133 hlsPlaylist: {
134 videoFilename: string
135 }
136 }
137
138 interface QuickTranscodeOptions extends BaseTranscodeOptions {
139 type: 'quick-transcode'
140 }
141
142 interface VideoTranscodeOptions extends BaseTranscodeOptions {
143 type: 'video'
144 }
145
146 interface MergeAudioTranscodeOptions extends BaseTranscodeOptions {
147 type: 'merge-audio'
148 audioPath: string
149 }
150
151 type TranscodeOptions = HLSTranscodeOptions | VideoTranscodeOptions | MergeAudioTranscodeOptions | QuickTranscodeOptions
152
153 function transcode (options: TranscodeOptions) {
154 return new Promise<void>(async (res, rej) => {
155 try {
156 let command = ffmpeg(options.inputPath, { niceness: FFMPEG_NICE.TRANSCODING })
157 .output(options.outputPath)
158
159 if (options.type === 'quick-transcode') {
160 command = await buildQuickTranscodeCommand(command)
161 } else if (options.type === 'hls') {
162 command = await buildHLSCommand(command, options)
163 } else if (options.type === 'merge-audio') {
164 command = await buildAudioMergeCommand(command, options)
165 } else {
166 command = await buildx264Command(command, options)
167 }
168
169 if (CONFIG.TRANSCODING.THREADS > 0) {
170 // if we don't set any threads ffmpeg will chose automatically
171 command = command.outputOption('-threads ' + CONFIG.TRANSCODING.THREADS)
172 }
173
174 command
175 .on('error', (err, stdout, stderr) => {
176 logger.error('Error in transcoding job.', { stdout, stderr })
177 return rej(err)
178 })
179 .on('end', () => {
180 return fixHLSPlaylistIfNeeded(options)
181 .then(() => res())
182 .catch(err => rej(err))
183 })
184 .run()
185 } catch (err) {
186 return rej(err)
187 }
188 })
189 }
190
191 async function canDoQuickTranscode (path: string): Promise<boolean> {
192 // NOTE: This could be optimized by running ffprobe only once (but it runs fast anyway)
193 const videoStream = await getVideoStreamFromFile(path)
194 const parsedAudio = await audio.get(path)
195 const fps = await getVideoFileFPS(path)
196 const bitRate = await getVideoFileBitrate(path)
197 const resolution = await getVideoFileResolution(path)
198
199 // check video params
200 if (videoStream[ 'codec_name' ] !== 'h264') return false
201 if (videoStream[ 'pix_fmt' ] !== 'yuv420p') return false
202 if (fps < VIDEO_TRANSCODING_FPS.MIN || fps > VIDEO_TRANSCODING_FPS.MAX) return false
203 if (bitRate > getMaxBitrate(resolution.videoFileResolution, fps, VIDEO_TRANSCODING_FPS)) return false
204
205 // check audio params (if audio stream exists)
206 if (parsedAudio.audioStream) {
207 if (parsedAudio.audioStream[ 'codec_name' ] !== 'aac') return false
208
209 const maxAudioBitrate = audio.bitrate[ 'aac' ](parsedAudio.audioStream[ 'bit_rate' ])
210 if (maxAudioBitrate !== -1 && parsedAudio.audioStream[ 'bit_rate' ] > maxAudioBitrate) return false
211 }
212
213 return true
214 }
215
216 // ---------------------------------------------------------------------------
217
218 export {
219 getVideoFileSize,
220 getVideoFileResolution,
221 getDurationFromVideoFile,
222 generateImageFromVideoFile,
223 TranscodeOptions,
224 TranscodeOptionsType,
225 transcode,
226 getVideoFileFPS,
227 computeResolutionsToTranscode,
228 audio,
229 getVideoFileBitrate,
230 canDoQuickTranscode
231 }
232
233 // ---------------------------------------------------------------------------
234
235 async function buildx264Command (command: ffmpeg.FfmpegCommand, options: VideoTranscodeOptions) {
236 let fps = await getVideoFileFPS(options.inputPath)
237 // On small/medium resolutions, limit FPS
238 if (
239 options.resolution !== undefined &&
240 options.resolution < VIDEO_TRANSCODING_FPS.KEEP_ORIGIN_FPS_RESOLUTION_MIN &&
241 fps > VIDEO_TRANSCODING_FPS.AVERAGE
242 ) {
243 fps = VIDEO_TRANSCODING_FPS.AVERAGE
244 }
245
246 command = await presetH264(command, options.inputPath, options.resolution, fps)
247
248 if (options.resolution !== undefined) {
249 // '?x720' or '720x?' for example
250 const size = options.isPortraitMode === true ? `${options.resolution}x?` : `?x${options.resolution}`
251 command = command.size(size)
252 }
253
254 if (fps) {
255 // Hard FPS limits
256 if (fps > VIDEO_TRANSCODING_FPS.MAX) fps = VIDEO_TRANSCODING_FPS.MAX
257 else if (fps < VIDEO_TRANSCODING_FPS.MIN) fps = VIDEO_TRANSCODING_FPS.MIN
258
259 command = command.withFPS(fps)
260 }
261
262 return command
263 }
264
265 async function buildAudioMergeCommand (command: ffmpeg.FfmpegCommand, options: MergeAudioTranscodeOptions) {
266 command = command.loop(undefined)
267
268 command = await presetH264VeryFast(command, options.audioPath, options.resolution)
269
270 command = command.input(options.audioPath)
271 .videoFilter('scale=trunc(iw/2)*2:trunc(ih/2)*2') // Avoid "height not divisible by 2" error
272 .outputOption('-tune stillimage')
273 .outputOption('-shortest')
274
275 return command
276 }
277
278 async function buildQuickTranscodeCommand (command: ffmpeg.FfmpegCommand) {
279 command = await presetCopy(command)
280
281 command = command.outputOption('-map_metadata -1') // strip all metadata
282 .outputOption('-movflags faststart')
283
284 return command
285 }
286
287 async function buildHLSCommand (command: ffmpeg.FfmpegCommand, options: HLSTranscodeOptions) {
288 const videoPath = getHLSVideoPath(options)
289
290 command = await presetCopy(command)
291
292 command = command.outputOption('-hls_time 4')
293 .outputOption('-hls_list_size 0')
294 .outputOption('-hls_playlist_type vod')
295 .outputOption('-hls_segment_filename ' + videoPath)
296 .outputOption('-hls_segment_type fmp4')
297 .outputOption('-f hls')
298 .outputOption('-hls_flags single_file')
299
300 return command
301 }
302
303 function getHLSVideoPath (options: HLSTranscodeOptions) {
304 return `${dirname(options.outputPath)}/${options.hlsPlaylist.videoFilename}`
305 }
306
307 async function fixHLSPlaylistIfNeeded (options: TranscodeOptions) {
308 if (options.type !== 'hls') return
309
310 const fileContent = await readFile(options.outputPath)
311
312 const videoFileName = options.hlsPlaylist.videoFilename
313 const videoFilePath = getHLSVideoPath(options)
314
315 // Fix wrong mapping with some ffmpeg versions
316 const newContent = fileContent.toString()
317 .replace(`#EXT-X-MAP:URI="${videoFilePath}",`, `#EXT-X-MAP:URI="${videoFileName}",`)
318
319 await writeFile(options.outputPath, newContent)
320 }
321
322 function getVideoStreamFromFile (path: string) {
323 return new Promise<any>((res, rej) => {
324 ffmpeg.ffprobe(path, (err, metadata) => {
325 if (err) return rej(err)
326
327 const videoStream = metadata.streams.find(s => s.codec_type === 'video')
328 if (!videoStream) return rej(new Error('Cannot find video stream of ' + path))
329
330 return res(videoStream)
331 })
332 })
333 }
334
335 /**
336 * A slightly customised version of the 'veryfast' x264 preset
337 *
338 * The veryfast preset is right in the sweet spot of performance
339 * and quality. Superfast and ultrafast will give you better
340 * performance, but then quality is noticeably worse.
341 */
342 async function presetH264VeryFast (command: ffmpeg.FfmpegCommand, input: string, resolution: VideoResolution, fps?: number) {
343 let localCommand = await presetH264(command, input, resolution, fps)
344
345 localCommand = localCommand.outputOption('-preset:v veryfast')
346
347 /*
348 MAIN reference: https://slhck.info/video/2017/03/01/rate-control.html
349 Our target situation is closer to a livestream than a stream,
350 since we want to reduce as much a possible the encoding burden,
351 although not to the point of a livestream where there is a hard
352 constraint on the frames per second to be encoded.
353 */
354
355 return localCommand
356 }
357
358 /**
359 * A toolbox to play with audio
360 */
361 namespace audio {
362 export const get = (option: string) => {
363 // without position, ffprobe considers the last input only
364 // we make it consider the first input only
365 // if you pass a file path to pos, then ffprobe acts on that file directly
366 return new Promise<{ absolutePath: string, audioStream?: any }>((res, rej) => {
367
368 function parseFfprobe (err: any, data: ffmpeg.FfprobeData) {
369 if (err) return rej(err)
370
371 if ('streams' in data) {
372 const audioStream = data.streams.find(stream => stream['codec_type'] === 'audio')
373 if (audioStream) {
374 return res({
375 absolutePath: data.format.filename,
376 audioStream
377 })
378 }
379 }
380
381 return res({ absolutePath: data.format.filename })
382 }
383
384 return ffmpeg.ffprobe(option, parseFfprobe)
385 })
386 }
387
388 export namespace bitrate {
389 const baseKbitrate = 384
390
391 const toBits = (kbits: number) => kbits * 8000
392
393 export const aac = (bitrate: number): number => {
394 switch (true) {
395 case bitrate > toBits(baseKbitrate):
396 return baseKbitrate
397
398 default:
399 return -1 // we interpret it as a signal to copy the audio stream as is
400 }
401 }
402
403 export const mp3 = (bitrate: number): number => {
404 /*
405 a 192kbit/sec mp3 doesn't hold as much information as a 192kbit/sec aac.
406 That's why, when using aac, we can go to lower kbit/sec. The equivalences
407 made here are not made to be accurate, especially with good mp3 encoders.
408 */
409 switch (true) {
410 case bitrate <= toBits(192):
411 return 128
412
413 case bitrate <= toBits(384):
414 return 256
415
416 default:
417 return baseKbitrate
418 }
419 }
420 }
421 }
422
423 /**
424 * Standard profile, with variable bitrate audio and faststart.
425 *
426 * As for the audio, quality '5' is the highest and ensures 96-112kbps/channel
427 * See https://trac.ffmpeg.org/wiki/Encode/AAC#fdk_vbr
428 */
429 async function presetH264 (command: ffmpeg.FfmpegCommand, input: string, resolution: VideoResolution, fps?: number) {
430 let localCommand = command
431 .format('mp4')
432 .videoCodec('libx264')
433 .outputOption('-level 3.1') // 3.1 is the minimal ressource allocation for our highest supported resolution
434 .outputOption('-b_strategy 1') // NOTE: b-strategy 1 - heuristic algorythm, 16 is optimal B-frames for it
435 .outputOption('-bf 16') // NOTE: Why 16: https://github.com/Chocobozzz/PeerTube/pull/774. b-strategy 2 -> B-frames<16
436 .outputOption('-pix_fmt yuv420p') // allows import of source material with incompatible pixel formats (e.g. MJPEG video)
437 .outputOption('-map_metadata -1') // strip all metadata
438 .outputOption('-movflags faststart')
439
440 const parsedAudio = await audio.get(input)
441
442 if (!parsedAudio.audioStream) {
443 localCommand = localCommand.noAudio()
444 } else if ((await checkFFmpegEncoders()).get('libfdk_aac')) { // we favor VBR, if a good AAC encoder is available
445 localCommand = localCommand
446 .audioCodec('libfdk_aac')
447 .audioQuality(5)
448 } else {
449 // we try to reduce the ceiling bitrate by making rough matches of bitrates
450 // of course this is far from perfect, but it might save some space in the end
451 localCommand = localCommand.audioCodec('aac')
452
453 const audioCodecName = parsedAudio.audioStream[ 'codec_name' ]
454
455 if (audio.bitrate[ audioCodecName ]) {
456 const bitrate = audio.bitrate[ audioCodecName ](parsedAudio.audioStream[ 'bit_rate' ])
457 if (bitrate !== undefined && bitrate !== -1) localCommand = localCommand.audioBitrate(bitrate)
458 }
459 }
460
461 if (fps) {
462 // Constrained Encoding (VBV)
463 // https://slhck.info/video/2017/03/01/rate-control.html
464 // https://trac.ffmpeg.org/wiki/Limiting%20the%20output%20bitrate
465 const targetBitrate = getTargetBitrate(resolution, fps, VIDEO_TRANSCODING_FPS)
466 localCommand = localCommand.outputOptions([ `-maxrate ${targetBitrate}`, `-bufsize ${targetBitrate * 2}` ])
467
468 // Keyframe interval of 2 seconds for faster seeking and resolution switching.
469 // https://streaminglearningcenter.com/blogs/whats-the-right-keyframe-interval.html
470 // https://superuser.com/a/908325
471 localCommand = localCommand.outputOption(`-g ${fps * 2}`)
472 }
473
474 return localCommand
475 }
476
477 async function presetCopy (command: ffmpeg.FfmpegCommand): Promise<ffmpeg.FfmpegCommand> {
478 return command
479 .format('mp4')
480 .videoCodec('copy')
481 .audioCodec('copy')
482 }