Add audio support in upload

17 files changed, 287 insertions, 149 deletions

diff --git a/server/assets/default-audio-background.jpg b/server/assets/default-audio-background.jpg
new file mode 100644
index 000000000..a19173eac
--- /dev/null
+++ b/server/assets/default-audio-background.jpg
diff --git a/server/controllers/api/config.ts b/server/controllers/api/config.ts
index 40012c03b..d9ce6a153 100644
--- a/server/controllers/api/config.ts
+++ b/server/controllers/api/config.ts
@@ -255,6 +255,7 @@ function customConfig (): CustomConfig {
     transcoding: {
       enabled: CONFIG.TRANSCODING.ENABLED,
       allowAdditionalExtensions: CONFIG.TRANSCODING.ALLOW_ADDITIONAL_EXTENSIONS,
+      allowAudioFiles: CONFIG.TRANSCODING.ALLOW_AUDIO_FILES,
       threads: CONFIG.TRANSCODING.THREADS,
       resolutions: {
         '240p': CONFIG.TRANSCODING.RESOLUTIONS[ '240p' ],
diff --git a/server/controllers/api/videos/index.ts b/server/controllers/api/videos/index.ts
index 1a18a8ae8..a2a615a79 100644
--- a/server/controllers/api/videos/index.ts
+++ b/server/controllers/api/videos/index.ts
@@ -1,12 +1,12 @@
 import * as express from 'express'
 import { extname, join } from 'path'
-import { VideoCreate, VideoPrivacy, VideoState, VideoUpdate } from '../../../../shared'
+import { VideoCreate, VideoPrivacy, VideoResolution, VideoState, VideoUpdate } from '../../../../shared'
 import { getVideoFileFPS, getVideoFileResolution } from '../../../helpers/ffmpeg-utils'
 import { logger } from '../../../helpers/logger'
 import { auditLoggerFactory, getAuditIdFromRes, VideoAuditView } from '../../../helpers/audit-logger'
 import { getFormattedObjects, getServerActor } from '../../../helpers/utils'
 import { autoBlacklistVideoIfNeeded } from '../../../lib/video-blacklist'
-import { MIMETYPES, VIDEO_CATEGORIES, VIDEO_LANGUAGES, VIDEO_LICENCES, VIDEO_PRIVACIES } from '../../../initializers/constants'
+import { MIMETYPES, VIDEO_CATEGORIES, VIDEO_LANGUAGES, VIDEO_LICENCES, VIDEO_PRIVACIES, DEFAULT_AUDIO_RESOLUTION } from '../../../initializers/constants'
 import {
   changeVideoChannelShare,
   federateVideoIfNeeded,
@@ -54,6 +54,7 @@ import { CONFIG } from '../../../initializers/config'
 import { sequelizeTypescript } from '../../../initializers/database'
 import { createVideoMiniatureFromExisting, generateVideoMiniature } from '../../../lib/thumbnail'
 import { ThumbnailType } from '../../../../shared/models/videos/thumbnail.type'
+import { VideoTranscodingPayload } from '../../../lib/job-queue/handlers/video-transcoding'
 
 const auditLogger = auditLoggerFactory('videos')
 const videosRouter = express.Router()
@@ -191,18 +192,19 @@ async function addVideo (req: express.Request, res: express.Response) {
   const video = new VideoModel(videoData)
   video.url = getVideoActivityPubUrl(video) // We use the UUID, so set the URL after building the object
 
-  // Build the file object
-  const { videoFileResolution } = await getVideoFileResolution(videoPhysicalFile.path)
-  const fps = await getVideoFileFPS(videoPhysicalFile.path)
-
   const videoFileData = {
     extname: extname(videoPhysicalFile.filename),
-    resolution: videoFileResolution,
-    size: videoPhysicalFile.size,
-    fps
+    size: videoPhysicalFile.size
   }
   const videoFile = new VideoFileModel(videoFileData)
 
+  if (!videoFile.isAudio()) {
+    videoFile.fps = await getVideoFileFPS(videoPhysicalFile.path)
+    videoFile.resolution = (await getVideoFileResolution(videoPhysicalFile.path)).videoFileResolution
+  } else {
+    videoFile.resolution = DEFAULT_AUDIO_RESOLUTION
+  }
+
   // Move physical file
   const videoDir = CONFIG.STORAGE.VIDEOS_DIR
   const destination = join(videoDir, video.getVideoFilename(videoFile))
@@ -279,9 +281,21 @@ async function addVideo (req: express.Request, res: express.Response) {
 
   if (video.state === VideoState.TO_TRANSCODE) {
     // Put uuid because we don't have id auto incremented for now
-    const dataInput = {
-      videoUUID: videoCreated.uuid,
-      isNewVideo: true
+    let dataInput: VideoTranscodingPayload
+
+    if (videoFile.isAudio()) {
+      dataInput = {
+        type: 'merge-audio' as 'merge-audio',
+        resolution: DEFAULT_AUDIO_RESOLUTION,
+        videoUUID: videoCreated.uuid,
+        isNewVideo: true
+      }
+    } else {
+      dataInput = {
+        type: 'optimize' as 'optimize',
+        videoUUID: videoCreated.uuid,
+        isNewVideo: true
+      }
     }
 
     await JobQueue.Instance.createJob({ type: 'video-transcoding', payload: dataInput })
diff --git a/server/controllers/static.ts b/server/controllers/static.ts
index 05019fcc2..d57dba6ce 100644
--- a/server/controllers/static.ts
+++ b/server/controllers/static.ts
@@ -181,7 +181,7 @@ async function getVideoCaption (req: express.Request, res: express.Response) {
   return res.sendFile(result.path, { maxAge: STATIC_MAX_AGE })
 }
 
-async function generateNodeinfo (req: express.Request, res: express.Response, next: express.NextFunction) {
+async function generateNodeinfo (req: express.Request, res: express.Response) {
   const { totalVideos } = await VideoModel.getStats()
   const { totalLocalVideoComments } = await VideoCommentModel.getStats()
   const { totalUsers } = await UserModel.getStats()
diff --git a/server/helpers/ffmpeg-utils.ts b/server/helpers/ffmpeg-utils.ts
index 2fdf34cb7..c180da832 100644
--- a/server/helpers/ffmpeg-utils.ts
+++ b/server/helpers/ffmpeg-utils.ts
@@ -117,37 +117,50 @@ async function generateImageFromVideoFile (fromPath: string, folder: string, ima
   }
 }
 
-type TranscodeOptions = {
+type TranscodeOptionsType = 'hls' | 'quick-transcode' | 'video' | 'merge-audio'
+
+interface BaseTranscodeOptions {
+  type: TranscodeOptionsType
   inputPath: string
   outputPath: string
   resolution: VideoResolution
   isPortraitMode?: boolean
-  doQuickTranscode?: Boolean
+}
 
-  hlsPlaylist?: {
+interface HLSTranscodeOptions extends BaseTranscodeOptions {
+  type: 'hls'
+  hlsPlaylist: {
     videoFilename: string
   }
 }
 
+interface QuickTranscodeOptions extends BaseTranscodeOptions {
+  type: 'quick-transcode'
+}
+
+interface VideoTranscodeOptions extends BaseTranscodeOptions {
+  type: 'video'
+}
+
+interface MergeAudioTranscodeOptions extends BaseTranscodeOptions {
+  type: 'merge-audio'
+  audioPath: string
+}
+
+type TranscodeOptions = HLSTranscodeOptions | VideoTranscodeOptions | MergeAudioTranscodeOptions | QuickTranscodeOptions
+
 function transcode (options: TranscodeOptions) {
   return new Promise<void>(async (res, rej) => {
     try {
       let command = ffmpeg(options.inputPath, { niceness: FFMPEG_NICE.TRANSCODING })
         .output(options.outputPath)
 
-      if (options.doQuickTranscode) {
-        if (options.hlsPlaylist) {
-          throw(Error("Quick transcode and HLS can't be used at the same time"))
-        }
-
-        command
-          .format('mp4')
-          .addOption('-c:v copy')
-          .addOption('-c:a copy')
-          .outputOption('-map_metadata -1') // strip all metadata
-          .outputOption('-movflags faststart')
-      } else if (options.hlsPlaylist) {
+      if (options.type === 'quick-transcode') {
+        command = await buildQuickTranscodeCommand(command)
+      } else if (options.type === 'hls') {
         command = await buildHLSCommand(command, options)
+      } else if (options.type === 'merge-audio') {
+        command = await buildAudioMergeCommand(command, options)
       } else {
         command = await buildx264Command(command, options)
       }
@@ -163,7 +176,7 @@ function transcode (options: TranscodeOptions) {
           return rej(err)
         })
         .on('end', () => {
-          return onTranscodingSuccess(options)
+          return fixHLSPlaylistIfNeeded(options)
             .then(() => res())
             .catch(err => rej(err))
         })
@@ -205,6 +218,8 @@ export {
   getVideoFileResolution,
   getDurationFromVideoFile,
   generateImageFromVideoFile,
+  TranscodeOptions,
+  TranscodeOptionsType,
   transcode,
   getVideoFileFPS,
   computeResolutionsToTranscode,
@@ -215,7 +230,7 @@ export {
 
 // ---------------------------------------------------------------------------
 
-async function buildx264Command (command: ffmpeg.FfmpegCommand, options: TranscodeOptions) {
+async function buildx264Command (command: ffmpeg.FfmpegCommand, options: VideoTranscodeOptions) {
   let fps = await getVideoFileFPS(options.inputPath)
   // On small/medium resolutions, limit FPS
   if (
@@ -226,7 +241,7 @@ async function buildx264Command (command: ffmpeg.FfmpegCommand, options: Transco
     fps = VIDEO_TRANSCODING_FPS.AVERAGE
   }
 
-  command = await presetH264(command, options.resolution, fps)
+  command = await presetH264(command, options.inputPath, options.resolution, fps)
 
   if (options.resolution !== undefined) {
     // '?x720' or '720x?' for example
@@ -245,7 +260,29 @@ async function buildx264Command (command: ffmpeg.FfmpegCommand, options: Transco
   return command
 }
 
-async function buildHLSCommand (command: ffmpeg.FfmpegCommand, options: TranscodeOptions) {
+async function buildAudioMergeCommand (command: ffmpeg.FfmpegCommand, options: MergeAudioTranscodeOptions) {
+  command = command.loop(undefined)
+
+  command = await presetH264VeryFast(command, options.audioPath, options.resolution)
+
+  command = command.input(options.audioPath)
+                   .videoFilter('scale=trunc(iw/2)*2:trunc(ih/2)*2') // Avoid "height not divisible by 2" error
+                   .outputOption('-tune stillimage')
+                   .outputOption('-shortest')
+
+  return command
+}
+
+async function buildQuickTranscodeCommand (command: ffmpeg.FfmpegCommand) {
+  command = await presetCopy(command)
+
+  command = command.outputOption('-map_metadata -1') // strip all metadata
+                   .outputOption('-movflags faststart')
+
+  return command
+}
+
+async function buildHLSCommand (command: ffmpeg.FfmpegCommand, options: HLSTranscodeOptions) {
   const videoPath = getHLSVideoPath(options)
 
   command = await presetCopy(command)
@@ -261,19 +298,19 @@ async function buildHLSCommand (command: ffmpeg.FfmpegCommand, options: Transcod
   return command
 }
 
-function getHLSVideoPath (options: TranscodeOptions) {
+function getHLSVideoPath (options: HLSTranscodeOptions) {
   return `${dirname(options.outputPath)}/${options.hlsPlaylist.videoFilename}`
 }
 
-async function onTranscodingSuccess (options: TranscodeOptions) {
-  if (!options.hlsPlaylist) return
+async function fixHLSPlaylistIfNeeded (options: TranscodeOptions) {
+  if (options.type !== 'hls') return
 
-  // Fix wrong mapping with some ffmpeg versions
   const fileContent = await readFile(options.outputPath)
 
   const videoFileName = options.hlsPlaylist.videoFilename
   const videoFilePath = getHLSVideoPath(options)
 
+  // Fix wrong mapping with some ffmpeg versions
   const newContent = fileContent.toString()
                                 .replace(`#EXT-X-MAP:URI="${videoFilePath}",`, `#EXT-X-MAP:URI="${videoFileName}",`)
 
@@ -300,44 +337,27 @@ function getVideoStreamFromFile (path: string) {
  * and quality. Superfast and ultrafast will give you better
  * performance, but then quality is noticeably worse.
  */
-async function presetH264VeryFast (command: ffmpeg.FfmpegCommand, resolution: VideoResolution, fps: number): Promise<ffmpeg.FfmpegCommand> {
-  let localCommand = await presetH264(command, resolution, fps)
+async function presetH264VeryFast (command: ffmpeg.FfmpegCommand, input: string, resolution: VideoResolution, fps?: number) {
+  let localCommand = await presetH264(command, input, resolution, fps)
+
   localCommand = localCommand.outputOption('-preset:v veryfast')
-             .outputOption([ '--aq-mode=2', '--aq-strength=1.3' ])
+
   /*
   MAIN reference: https://slhck.info/video/2017/03/01/rate-control.html
   Our target situation is closer to a livestream than a stream,
   since we want to reduce as much a possible the encoding burden,
-  altough not to the point of a livestream where there is a hard
+  although not to the point of a livestream where there is a hard
   constraint on the frames per second to be encoded.
-
-  why '--aq-mode=2 --aq-strength=1.3' instead of '-profile:v main'?
-    Make up for most of the loss of grain and macroblocking
-    with less computing power.
   */
 
   return localCommand
 }
 
 /**
- * A preset optimised for a stillimage audio video
- */
-async function presetStillImageWithAudio (
-  command: ffmpeg.FfmpegCommand,
-  resolution: VideoResolution,
-  fps: number
-): Promise<ffmpeg.FfmpegCommand> {
-  let localCommand = await presetH264VeryFast(command, resolution, fps)
-  localCommand = localCommand.outputOption('-tune stillimage')
-
-  return localCommand
-}
-
-/**
  * A toolbox to play with audio
  */
 namespace audio {
-  export const get = (option: ffmpeg.FfmpegCommand | string) => {
+  export const get = (option: string) => {
     // without position, ffprobe considers the last input only
     // we make it consider the first input only
     // if you pass a file path to pos, then ffprobe acts on that file directly
@@ -359,11 +379,7 @@ namespace audio {
         return res({ absolutePath: data.format.filename })
       }
 
-      if (typeof option === 'string') {
-        return ffmpeg.ffprobe(option, parseFfprobe)
-      }
-
-      return option.ffprobe(parseFfprobe)
+      return ffmpeg.ffprobe(option, parseFfprobe)
     })
   }
 
@@ -405,7 +421,7 @@ namespace audio {
  * As for the audio, quality '5' is the highest and ensures 96-112kbps/channel
  * See https://trac.ffmpeg.org/wiki/Encode/AAC#fdk_vbr
  */
-async function presetH264 (command: ffmpeg.FfmpegCommand, resolution: VideoResolution, fps: number): Promise<ffmpeg.FfmpegCommand> {
+async function presetH264 (command: ffmpeg.FfmpegCommand, input: string, resolution: VideoResolution, fps?: number) {
   let localCommand = command
     .format('mp4')
     .videoCodec('libx264')
@@ -416,7 +432,7 @@ async function presetH264 (command: ffmpeg.FfmpegCommand, resolution: VideoResol
     .outputOption('-map_metadata -1') // strip all metadata
     .outputOption('-movflags faststart')
 
-  const parsedAudio = await audio.get(localCommand)
+  const parsedAudio = await audio.get(input)
 
   if (!parsedAudio.audioStream) {
     localCommand = localCommand.noAudio()
@@ -425,28 +441,30 @@ async function presetH264 (command: ffmpeg.FfmpegCommand, resolution: VideoResol
       .audioCodec('libfdk_aac')
       .audioQuality(5)
   } else {
-    // we try to reduce the ceiling bitrate by making rough correspondances of bitrates
+    // we try to reduce the ceiling bitrate by making rough matches of bitrates
     // of course this is far from perfect, but it might save some space in the end
+    localCommand = localCommand.audioCodec('aac')
+
     const audioCodecName = parsedAudio.audioStream[ 'codec_name' ]
-    let bitrate: number
-    if (audio.bitrate[ audioCodecName ]) {
-      localCommand = localCommand.audioCodec('aac')
 
-      bitrate = audio.bitrate[ audioCodecName ](parsedAudio.audioStream[ 'bit_rate' ])
+    if (audio.bitrate[ audioCodecName ]) {
+      const bitrate = audio.bitrate[ audioCodecName ](parsedAudio.audioStream[ 'bit_rate' ])
       if (bitrate !== undefined && bitrate !== -1) localCommand = localCommand.audioBitrate(bitrate)
     }
   }
 
-  // Constrained Encoding (VBV)
-  // https://slhck.info/video/2017/03/01/rate-control.html
-  // https://trac.ffmpeg.org/wiki/Limiting%20the%20output%20bitrate
-  const targetBitrate = getTargetBitrate(resolution, fps, VIDEO_TRANSCODING_FPS)
-  localCommand = localCommand.outputOptions([`-maxrate ${ targetBitrate }`, `-bufsize ${ targetBitrate * 2 }`])
-
-  // Keyframe interval of 2 seconds for faster seeking and resolution switching.
-  // https://streaminglearningcenter.com/blogs/whats-the-right-keyframe-interval.html
-  // https://superuser.com/a/908325
-  localCommand = localCommand.outputOption(`-g ${ fps * 2 }`)
+  if (fps) {
+    // Constrained Encoding (VBV)
+    // https://slhck.info/video/2017/03/01/rate-control.html
+    // https://trac.ffmpeg.org/wiki/Limiting%20the%20output%20bitrate
+    const targetBitrate = getTargetBitrate(resolution, fps, VIDEO_TRANSCODING_FPS)
+    localCommand = localCommand.outputOptions([ `-maxrate ${targetBitrate}`, `-bufsize ${targetBitrate * 2}` ])
+
+    // Keyframe interval of 2 seconds for faster seeking and resolution switching.
+    // https://streaminglearningcenter.com/blogs/whats-the-right-keyframe-interval.html
+    // https://superuser.com/a/908325
+    localCommand = localCommand.outputOption(`-g ${fps * 2}`)
+  }
 
   return localCommand
 }
diff --git a/server/initializers/config.ts b/server/initializers/config.ts
index 4f77e144d..4515bc804 100644
--- a/server/initializers/config.ts
+++ b/server/initializers/config.ts
@@ -140,6 +140,7 @@ const CONFIG = {
   TRANSCODING: {
     get ENABLED () { return config.get<boolean>('transcoding.enabled') },
     get ALLOW_ADDITIONAL_EXTENSIONS () { return config.get<boolean>('transcoding.allow_additional_extensions') },
+    get ALLOW_AUDIO_FILES () { return config.get<boolean>('transcoding.allow_audio_files') },
     get THREADS () { return config.get<number>('transcoding.threads') },
     RESOLUTIONS: {
       get '240p' () { return config.get<boolean>('transcoding.resolutions.240p') },
diff --git a/server/initializers/constants.ts b/server/initializers/constants.ts
index 62778ae58..718d0893b 100644
--- a/server/initializers/constants.ts
+++ b/server/initializers/constants.ts
@@ -1,10 +1,10 @@
 import { join } from 'path'
-import { JobType, VideoRateType, VideoState } from '../../shared/models'
+import { JobType, VideoRateType, VideoResolution, VideoState } from '../../shared/models'
 import { ActivityPubActorType } from '../../shared/models/activitypub'
 import { FollowState } from '../../shared/models/actors'
 import { VideoAbuseState, VideoImportState, VideoPrivacy, VideoTranscodingFPS } from '../../shared/models/videos'
 // Do not use barrels, remain constants as independent as possible
-import { isTestInstance, sanitizeHost, sanitizeUrl } from '../helpers/core-utils'
+import { isTestInstance, sanitizeHost, sanitizeUrl, root } from '../helpers/core-utils'
 import { NSFWPolicyType } from '../../shared/models/videos/nsfw-policy.type'
 import { invert } from 'lodash'
 import { CronRepeatOptions, EveryRepeatOptions } from 'bull'
@@ -228,7 +228,7 @@ let CONSTRAINTS_FIELDS = {
         max: 2 * 1024 * 1024 // 2MB
       }
     },
-    EXTNAME: buildVideosExtname(),
+    EXTNAME: [] as string[],
     INFO_HASH: { min: 40, max: 40 }, // Length, info hash is 20 bytes length but we represent it in hexadecimal so 20 * 2
     DURATION: { min: 0 }, // Number
     TAGS: { min: 0, max: 5 }, // Number of total tags
@@ -300,6 +300,8 @@ const VIDEO_TRANSCODING_FPS: VideoTranscodingFPS = {
   KEEP_ORIGIN_FPS_RESOLUTION_MIN: 720 // We keep the original FPS on high resolutions (720 minimum)
 }
 
+const DEFAULT_AUDIO_RESOLUTION = VideoResolution.H_480P
+
 const VIDEO_RATE_TYPES: { [ id: string ]: VideoRateType } = {
   LIKE: 'like',
   DISLIKE: 'dislike'
@@ -380,8 +382,18 @@ const VIDEO_PLAYLIST_TYPES = {
 }
 
 const MIMETYPES = {
+  AUDIO: {
+    MIMETYPE_EXT: {
+      'audio/mpeg': '.mp3',
+      'audio/mp3': '.mp3',
+      'application/ogg': '.ogg',
+      'audio/ogg': '.ogg',
+      'audio/flac': '.flac'
+    },
+    EXT_MIMETYPE: null as { [ id: string ]: string }
+  },
   VIDEO: {
-    MIMETYPE_EXT: buildVideoMimetypeExt(),
+    MIMETYPE_EXT: null as { [ id: string ]: string },
     EXT_MIMETYPE: null as { [ id: string ]: string }
   },
   IMAGE: {
@@ -403,7 +415,7 @@ const MIMETYPES = {
     }
   }
 }
-MIMETYPES.VIDEO.EXT_MIMETYPE = invert(MIMETYPES.VIDEO.MIMETYPE_EXT)
+MIMETYPES.AUDIO.EXT_MIMETYPE = invert(MIMETYPES.AUDIO.MIMETYPE_EXT)
 
 // ---------------------------------------------------------------------------
 
@@ -429,7 +441,7 @@ const ACTIVITY_PUB = {
   COLLECTION_ITEMS_PER_PAGE: 10,
   FETCH_PAGE_LIMIT: 100,
   URL_MIME_TYPES: {
-    VIDEO: Object.keys(MIMETYPES.VIDEO.MIMETYPE_EXT),
+    VIDEO: [] as string[],
     TORRENT: [ 'application/x-bittorrent' ],
     MAGNET: [ 'application/x-bittorrent;x-scheme-handler/magnet' ]
   },
@@ -543,6 +555,10 @@ const REDUNDANCY = {
 
 const ACCEPT_HEADERS = [ 'html', 'application/json' ].concat(ACTIVITY_PUB.POTENTIAL_ACCEPT_HEADERS)
 
+const ASSETS_PATH = {
+  DEFAULT_AUDIO_BACKGROUND: join(root(), 'server', 'assets', 'default-audio-background.jpg')
+}
+
 // ---------------------------------------------------------------------------
 
 const CUSTOM_HTML_TAG_COMMENTS = {
@@ -612,6 +628,7 @@ if (isTestInstance() === true) {
 }
 
 updateWebserverUrls()
+updateWebserverConfig()
 
 registerConfigChangedHandler(() => {
   updateWebserverUrls()
@@ -681,12 +698,14 @@ export {
   RATES_LIMIT,
   MIMETYPES,
   CRAWL_REQUEST_CONCURRENCY,
+  DEFAULT_AUDIO_RESOLUTION,
   JOB_COMPLETED_LIFETIME,
   HTTP_SIGNATURE,
   VIDEO_IMPORT_STATES,
   VIDEO_VIEW_LIFETIME,
   CONTACT_FORM_LIFETIME,
   VIDEO_PLAYLIST_PRIVACIES,
+  ASSETS_PATH,
   loadLanguages,
   buildLanguages
 }
@@ -700,15 +719,21 @@ function buildVideoMimetypeExt () {
     'video/mp4': '.mp4'
   }
 
-  if (CONFIG.TRANSCODING.ENABLED && CONFIG.TRANSCODING.ALLOW_ADDITIONAL_EXTENSIONS) {
-    Object.assign(data, {
-      'video/quicktime': '.mov',
-      'video/x-msvideo': '.avi',
-      'video/x-flv': '.flv',
-      'video/x-matroska': '.mkv',
-      'application/octet-stream': '.mkv',
-      'video/avi': '.avi'
-    })
+  if (CONFIG.TRANSCODING.ENABLED) {
+    if (CONFIG.TRANSCODING.ALLOW_ADDITIONAL_EXTENSIONS) {
+      Object.assign(data, {
+        'video/quicktime': '.mov',
+        'video/x-msvideo': '.avi',
+        'video/x-flv': '.flv',
+        'video/x-matroska': '.mkv',
+        'application/octet-stream': '.mkv',
+        'video/avi': '.avi'
+      })
+    }
+
+    if (CONFIG.TRANSCODING.ALLOW_AUDIO_FILES) {
+      Object.assign(data, MIMETYPES.AUDIO.MIMETYPE_EXT)
+    }
   }
 
   return data
@@ -724,16 +749,15 @@ function updateWebserverUrls () {
 }
 
 function updateWebserverConfig () {
-  CONSTRAINTS_FIELDS.VIDEOS.EXTNAME = buildVideosExtname()
-
   MIMETYPES.VIDEO.MIMETYPE_EXT = buildVideoMimetypeExt()
   MIMETYPES.VIDEO.EXT_MIMETYPE = invert(MIMETYPES.VIDEO.MIMETYPE_EXT)
+  ACTIVITY_PUB.URL_MIME_TYPES.VIDEO = Object.keys(MIMETYPES.VIDEO.MIMETYPE_EXT)
+
+  CONSTRAINTS_FIELDS.VIDEOS.EXTNAME = buildVideosExtname()
 }
 
 function buildVideosExtname () {
-  return CONFIG.TRANSCODING.ENABLED && CONFIG.TRANSCODING.ALLOW_ADDITIONAL_EXTENSIONS
-    ? [ '.mp4', '.ogv', '.webm', '.mkv', '.mov', '.avi', '.flv' ]
-    : [ '.mp4', '.ogv', '.webm' ]
+  return Object.keys(MIMETYPES.VIDEO.EXT_MIMETYPE)
 }
 
 function loadLanguages () {
diff --git a/server/lib/files-cache/videos-preview-cache.ts b/server/lib/files-cache/videos-preview-cache.ts
index 14be7f24a..a68619d07 100644
--- a/server/lib/files-cache/videos-preview-cache.ts
+++ b/server/lib/files-cache/videos-preview-cache.ts
@@ -21,7 +21,7 @@ class VideosPreviewCache extends AbstractVideoStaticFileCache <string> {
     const video = await VideoModel.loadByUUIDWithFile(videoUUID)
     if (!video) return undefined
 
-    if (video.isOwned()) return { isOwned: true, path: join(CONFIG.STORAGE.PREVIEWS_DIR, video.getPreview().filename) }
+    if (video.isOwned()) return { isOwned: true, path: video.getPreview().getPath() }
 
     return this.loadRemoteFile(videoUUID)
   }
diff --git a/server/lib/job-queue/handlers/video-file-import.ts b/server/lib/job-queue/handlers/video-file-import.ts
index 921d9a083..8cacb0ef3 100644
--- a/server/lib/job-queue/handlers/video-file-import.ts
+++ b/server/lib/job-queue/handlers/video-file-import.ts
@@ -1,7 +1,7 @@
 import * as Bull from 'bull'
 import { logger } from '../../../helpers/logger'
 import { VideoModel } from '../../../models/video/video'
-import { publishVideoIfNeeded } from './video-transcoding'
+import { publishNewResolutionIfNeeded } from './video-transcoding'
 import { getVideoFileFPS, getVideoFileResolution } from '../../../helpers/ffmpeg-utils'
 import { copy, stat } from 'fs-extra'
 import { VideoFileModel } from '../../../models/video/video-file'
@@ -25,7 +25,7 @@ async function processVideoFileImport (job: Bull.Job) {
 
   await updateVideoFile(video, payload.filePath)
 
-  await publishVideoIfNeeded(video)
+  await publishNewResolutionIfNeeded(video)
   return video
 }
 
diff --git a/server/lib/job-queue/handlers/video-import.ts b/server/lib/job-queue/handlers/video-import.ts
index 1650916a6..50e159245 100644
--- a/server/lib/job-queue/handlers/video-import.ts
+++ b/server/lib/job-queue/handlers/video-import.ts
@@ -209,6 +209,7 @@ async function processFile (downloader: () => Promise<string>, videoImport: Vide
     if (videoImportUpdated.Video.state === VideoState.TO_TRANSCODE) {
       // Put uuid because we don't have id auto incremented for now
       const dataInput = {
+        type: 'optimize' as 'optimize',
         videoUUID: videoImportUpdated.Video.uuid,
         isNewVideo: true
       }
diff --git a/server/lib/job-queue/handlers/video-transcoding.ts b/server/lib/job-queue/handlers/video-transcoding.ts
index 48cac517e..e9b84ecd6 100644
--- a/server/lib/job-queue/handlers/video-transcoding.ts
+++ b/server/lib/job-queue/handlers/video-transcoding.ts
@@ -8,18 +8,39 @@ import { retryTransactionWrapper } from '../../../helpers/database-utils'
 import { sequelizeTypescript } from '../../../initializers'
 import * as Bluebird from 'bluebird'
 import { computeResolutionsToTranscode } from '../../../helpers/ffmpeg-utils'
-import { generateHlsPlaylist, optimizeVideofile, transcodeOriginalVideofile } from '../../video-transcoding'
+import { generateHlsPlaylist, optimizeVideofile, transcodeOriginalVideofile, mergeAudioVideofile } from '../../video-transcoding'
 import { Notifier } from '../../notifier'
 import { CONFIG } from '../../../initializers/config'
 
-export type VideoTranscodingPayload = {
+interface BaseTranscodingPayload {
   videoUUID: string
-  resolution?: VideoResolution
   isNewVideo?: boolean
+}
+
+interface HLSTranscodingPayload extends BaseTranscodingPayload {
+  type: 'hls'
+  isPortraitMode?: boolean
+  resolution: VideoResolution
+}
+
+interface NewResolutionTranscodingPayload extends BaseTranscodingPayload {
+  type: 'new-resolution'
   isPortraitMode?: boolean
-  generateHlsPlaylist?: boolean
+  resolution: VideoResolution
+}
+
+interface MergeAudioTranscodingPayload extends BaseTranscodingPayload {
+  type: 'merge-audio'
+  resolution: VideoResolution
+}
+
+interface OptimizeTranscodingPayload extends BaseTranscodingPayload {
+  type: 'optimize'
 }
 
+export type VideoTranscodingPayload = HLSTranscodingPayload | NewResolutionTranscodingPayload
+  | OptimizeTranscodingPayload | MergeAudioTranscodingPayload
+
 async function processVideoTranscoding (job: Bull.Job) {
   const payload = job.data as VideoTranscodingPayload
   logger.info('Processing video file in job %d.', job.id)
@@ -31,14 +52,18 @@ async function processVideoTranscoding (job: Bull.Job) {
     return undefined
   }
 
-  if (payload.generateHlsPlaylist) {
+  if (payload.type === 'hls') {
     await generateHlsPlaylist(video, payload.resolution, payload.isPortraitMode || false)
 
     await retryTransactionWrapper(onHlsPlaylistGenerationSuccess, video)
-  } else if (payload.resolution) { // Transcoding in other resolution
+  } else if (payload.type === 'new-resolution') {
     await transcodeOriginalVideofile(video, payload.resolution, payload.isPortraitMode || false)
 
-    await retryTransactionWrapper(publishVideoIfNeeded, video, payload)
+    await retryTransactionWrapper(publishNewResolutionIfNeeded, video, payload)
+  } else if (payload.type === 'merge-audio') {
+    await mergeAudioVideofile(video, payload.resolution)
+
+    await retryTransactionWrapper(publishNewResolutionIfNeeded, video, payload)
   } else {
     await optimizeVideofile(video)
 
@@ -62,7 +87,7 @@ async function onHlsPlaylistGenerationSuccess (video: VideoModel) {
   })
 }
 
-async function publishVideoIfNeeded (video: VideoModel, payload?: VideoTranscodingPayload) {
+async function publishNewResolutionIfNeeded (video: VideoModel, payload?: NewResolutionTranscodingPayload | MergeAudioTranscodingPayload) {
   const { videoDatabase, videoPublished } = await sequelizeTypescript.transaction(async t => {
     // Maybe the video changed in database, refresh it
     let videoDatabase = await VideoModel.loadAndPopulateAccountAndServerAndTags(video.uuid, t)
@@ -94,7 +119,7 @@ async function publishVideoIfNeeded (video: VideoModel, payload?: VideoTranscodi
   await createHlsJobIfEnabled(payload)
 }
 
-async function onVideoFileOptimizerSuccess (videoArg: VideoModel, payload: VideoTranscodingPayload) {
+async function onVideoFileOptimizerSuccess (videoArg: VideoModel, payload: OptimizeTranscodingPayload) {
   if (videoArg === undefined) return undefined
 
   // Outside the transaction (IO on disk)
@@ -120,6 +145,7 @@ async function onVideoFileOptimizerSuccess (videoArg: VideoModel, payload: Video
 
       for (const resolution of resolutionsEnabled) {
         const dataInput = {
+          type: 'new-resolution' as 'new-resolution',
           videoUUID: videoDatabase.uuid,
           resolution
         }
@@ -149,27 +175,27 @@ async function onVideoFileOptimizerSuccess (videoArg: VideoModel, payload: Video
   if (payload.isNewVideo) Notifier.Instance.notifyOnNewVideo(videoDatabase)
   if (videoPublished) Notifier.Instance.notifyOnVideoPublishedAfterTranscoding(videoDatabase)
 
-  await createHlsJobIfEnabled(Object.assign({}, payload, { resolution: videoDatabase.getOriginalFile().resolution }))
+  const hlsPayload = Object.assign({}, payload, { resolution: videoDatabase.getOriginalFile().resolution })
+  await createHlsJobIfEnabled(hlsPayload)
 }
 
 // ---------------------------------------------------------------------------
 
 export {
   processVideoTranscoding,
-  publishVideoIfNeeded
+  publishNewResolutionIfNeeded
 }
 
 // ---------------------------------------------------------------------------
 
-function createHlsJobIfEnabled (payload?: VideoTranscodingPayload) {
+function createHlsJobIfEnabled (payload?: { videoUUID: string, resolution: number, isPortraitMode?: boolean }) {
   // Generate HLS playlist?
   if (payload && CONFIG.TRANSCODING.HLS.ENABLED) {
     const hlsTranscodingPayload = {
+      type: 'hls' as 'hls',
       videoUUID: payload.videoUUID,
       resolution: payload.resolution,
-      isPortraitMode: payload.isPortraitMode,
-
-      generateHlsPlaylist: true
+      isPortraitMode: payload.isPortraitMode
     }
 
     return JobQueue.Instance.createJob({ type: 'video-transcoding', payload: hlsTranscodingPayload })
diff --git a/server/lib/thumbnail.ts b/server/lib/thumbnail.ts
index 950b14c3b..18bdcded4 100644
--- a/server/lib/thumbnail.ts
+++ b/server/lib/thumbnail.ts
@@ -1,7 +1,7 @@
 import { VideoFileModel } from '../models/video/video-file'
 import { generateImageFromVideoFile } from '../helpers/ffmpeg-utils'
 import { CONFIG } from '../initializers/config'
-import { PREVIEWS_SIZE, THUMBNAILS_SIZE } from '../initializers/constants'
+import { PREVIEWS_SIZE, THUMBNAILS_SIZE, ASSETS_PATH } from '../initializers/constants'
 import { VideoModel } from '../models/video/video'
 import { ThumbnailModel } from '../models/video/thumbnail'
 import { ThumbnailType } from '../../shared/models/videos/thumbnail.type'
@@ -45,8 +45,10 @@ function createVideoMiniatureFromExisting (inputPath: string, video: VideoModel,
 function generateVideoMiniature (video: VideoModel, videoFile: VideoFileModel, type: ThumbnailType) {
   const input = video.getVideoFilePath(videoFile)
 
-  const { filename, basePath, height, width, existingThumbnail } = buildMetadataFromVideo(video, type)
-  const thumbnailCreator = () => generateImageFromVideoFile(input, basePath, filename, { height, width })
+  const { filename, basePath, height, width, existingThumbnail, outputPath } = buildMetadataFromVideo(video, type)
+  const thumbnailCreator = videoFile.isAudio()
+    ? () => processImage(ASSETS_PATH.DEFAULT_AUDIO_BACKGROUND, outputPath, { width, height }, true)
+    : () => generateImageFromVideoFile(input, basePath, filename, { height, width })
 
   return createThumbnailFromFunction({ thumbnailCreator, filename, height, width, type, existingThumbnail })
 }
diff --git a/server/lib/video-transcoding.ts b/server/lib/video-transcoding.ts
index d6b6b251a..8d786e0ef 100644
--- a/server/lib/video-transcoding.ts
+++ b/server/lib/video-transcoding.ts
@@ -1,6 +1,6 @@
 import { HLS_STREAMING_PLAYLIST_DIRECTORY, P2P_MEDIA_LOADER_PEER_VERSION, WEBSERVER } from '../initializers/constants'
 import { join } from 'path'
-import { getVideoFileFPS, transcode, canDoQuickTranscode } from '../helpers/ffmpeg-utils'
+import { canDoQuickTranscode, getVideoFileFPS, transcode, TranscodeOptions, TranscodeOptionsType } from '../helpers/ffmpeg-utils'
 import { ensureDir, move, remove, stat } from 'fs-extra'
 import { logger } from '../helpers/logger'
 import { VideoResolution } from '../../shared/models/videos'
@@ -23,13 +23,15 @@ async function optimizeVideofile (video: VideoModel, inputVideoFileArg?: VideoFi
   const videoInputPath = join(videosDirectory, video.getVideoFilename(inputVideoFile))
   const videoTranscodedPath = join(transcodeDirectory, video.id + '-transcoded' + newExtname)
 
-  const doQuickTranscode = await(canDoQuickTranscode(videoInputPath))
+  const transcodeType: TranscodeOptionsType = await canDoQuickTranscode(videoInputPath)
+    ? 'quick-transcode'
+    : 'video'
 
-  const transcodeOptions = {
+  const transcodeOptions: TranscodeOptions = {
+    type: transcodeType as any, // FIXME: typing issue
     inputPath: videoInputPath,
     outputPath: videoTranscodedPath,
-    resolution: inputVideoFile.resolution,
-    doQuickTranscode
+    resolution: inputVideoFile.resolution
   }
 
   // Could be very long!
@@ -39,19 +41,11 @@ async function optimizeVideofile (video: VideoModel, inputVideoFileArg?: VideoFi
     await remove(videoInputPath)
 
     // Important to do this before getVideoFilename() to take in account the new file extension
-    inputVideoFile.set('extname', newExtname)
-
-    const stats = await stat(videoTranscodedPath)
-    const fps = await getVideoFileFPS(videoTranscodedPath)
+    inputVideoFile.extname = newExtname
 
     const videoOutputPath = video.getVideoFilePath(inputVideoFile)
-    await move(videoTranscodedPath, videoOutputPath)
 
-    inputVideoFile.set('size', stats.size)
-    inputVideoFile.set('fps', fps)
-
-    await video.createTorrentAndSetInfoHash(inputVideoFile)
-    await inputVideoFile.save()
+    await onVideoFileTranscoding(video, inputVideoFile, videoTranscodedPath, videoOutputPath)
   } catch (err) {
     // Auto destruction...
     video.destroy().catch(err => logger.error('Cannot destruct video after transcoding failure.', { err }))
@@ -81,6 +75,7 @@ async function transcodeOriginalVideofile (video: VideoModel, resolution: VideoR
   const videoTranscodedPath = join(transcodeDirectory, video.getVideoFilename(newVideoFile))
 
   const transcodeOptions = {
+    type: 'video' as 'video',
     inputPath: videoInputPath,
     outputPath: videoTranscodedPath,
     resolution,
@@ -89,19 +84,37 @@ async function transcodeOriginalVideofile (video: VideoModel, resolution: VideoR
 
   await transcode(transcodeOptions)
 
-  const stats = await stat(videoTranscodedPath)
-  const fps = await getVideoFileFPS(videoTranscodedPath)
+  return onVideoFileTranscoding(video, newVideoFile, videoTranscodedPath, videoOutputPath)
+}
+
+async function mergeAudioVideofile (video: VideoModel, resolution: VideoResolution) {
+  const videosDirectory = CONFIG.STORAGE.VIDEOS_DIR
+  const transcodeDirectory = CONFIG.STORAGE.TMP_DIR
+  const newExtname = '.mp4'
+
+  const inputVideoFile = video.getOriginalFile()
 
-  await move(videoTranscodedPath, videoOutputPath)
+  const audioInputPath = join(videosDirectory, video.getVideoFilename(video.getOriginalFile()))
+  const videoTranscodedPath = join(transcodeDirectory, video.id + '-transcoded' + newExtname)
 
-  newVideoFile.set('size', stats.size)
-  newVideoFile.set('fps', fps)
+  const transcodeOptions = {
+    type: 'merge-audio' as 'merge-audio',
+    inputPath: video.getPreview().getPath(),
+    outputPath: videoTranscodedPath,
+    audioPath: audioInputPath,
+    resolution
+  }
 
-  await video.createTorrentAndSetInfoHash(newVideoFile)
+  await transcode(transcodeOptions)
 
-  await newVideoFile.save()
+  await remove(audioInputPath)
 
-  video.VideoFiles.push(newVideoFile)
+  // Important to do this before getVideoFilename() to take in account the new file extension
+  inputVideoFile.extname = newExtname
+
+  const videoOutputPath = video.getVideoFilePath(inputVideoFile)
+
+  return onVideoFileTranscoding(video, inputVideoFile, videoTranscodedPath, videoOutputPath)
 }
 
 async function generateHlsPlaylist (video: VideoModel, resolution: VideoResolution, isPortraitMode: boolean) {
@@ -112,6 +125,7 @@ async function generateHlsPlaylist (video: VideoModel, resolution: VideoResoluti
   const outputPath = join(baseHlsDirectory, VideoStreamingPlaylistModel.getHlsPlaylistFilename(resolution))
 
   const transcodeOptions = {
+    type: 'hls' as 'hls',
     inputPath: videoInputPath,
     outputPath,
     resolution,
@@ -140,8 +154,34 @@ async function generateHlsPlaylist (video: VideoModel, resolution: VideoResoluti
   })
 }
 
+// ---------------------------------------------------------------------------
+
 export {
   generateHlsPlaylist,
   optimizeVideofile,
-  transcodeOriginalVideofile
+  transcodeOriginalVideofile,
+  mergeAudioVideofile
+}
+
+// ---------------------------------------------------------------------------
+
+async function onVideoFileTranscoding (video: VideoModel, videoFile: VideoFileModel, transcodingPath: string, outputPath: string) {
+  const stats = await stat(transcodingPath)
+  const fps = await getVideoFileFPS(transcodingPath)
+
+  await move(transcodingPath, outputPath)
+
+  videoFile.set('size', stats.size)
+  videoFile.set('fps', fps)
+
+  await video.createTorrentAndSetInfoHash(videoFile)
+
+  const updatedVideoFile = await videoFile.save()
+
+  // Add it if this is a new created file
+  if (video.VideoFiles.some(f => f.id === videoFile.id) === false) {
+    video.VideoFiles.push(updatedVideoFile)
+  }
+
+  return video
 }
diff --git a/server/models/video/thumbnail.ts b/server/models/video/thumbnail.ts
index 206e9a3d6..8faf0adba 100644
--- a/server/models/video/thumbnail.ts
+++ b/server/models/video/thumbnail.ts
@@ -107,10 +107,12 @@ export class ThumbnailModel extends Model<ThumbnailModel> {
     return WEBSERVER.URL + staticPath + this.filename
   }
 
-  removeThumbnail () {
+  getPath () {
     const directory = ThumbnailModel.types[this.type].directory
-    const thumbnailPath = join(directory, this.filename)
+    return join(directory, this.filename)
+  }
 
-    return remove(thumbnailPath)
+  removeThumbnail () {
+    return remove(this.getPath())
   }
 }
diff --git a/server/models/video/video-file.ts b/server/models/video/video-file.ts
index 2203a7aba..05c490759 100644
--- a/server/models/video/video-file.ts
+++ b/server/models/video/video-file.ts
@@ -24,6 +24,7 @@ import { VideoModel } from './video'
 import { VideoRedundancyModel } from '../redundancy/video-redundancy'
 import { VideoStreamingPlaylistModel } from './video-streaming-playlist'
 import { FindOptions, QueryTypes, Transaction } from 'sequelize'
+import { MIMETYPES } from '../../initializers/constants'
 
 @Table({
   tableName: 'videoFile',
@@ -161,6 +162,10 @@ export class VideoFileModel extends Model<VideoFileModel> {
       }))
   }
 
+  isAudio () {
+    return !!MIMETYPES.AUDIO.EXT_MIMETYPE[this.extname]
+  }
+
   hasSameUniqueKeysThan (other: VideoFileModel) {
     return this.fps === other.fps &&
       this.resolution === other.resolution &&
diff --git a/server/tests/api/check-params/config.ts b/server/tests/api/check-params/config.ts
index 2a2ec606a..8155e11ab 100644
--- a/server/tests/api/check-params/config.ts
+++ b/server/tests/api/check-params/config.ts
@@ -59,6 +59,7 @@ describe('Test config API validators', function () {
     transcoding: {
       enabled: true,
       allowAdditionalExtensions: true,
+      allowAudioFiles: true,
       threads: 1,
       resolutions: {
         '240p': false,
diff --git a/server/tests/api/server/config.ts b/server/tests/api/server/config.ts
index ca389b7b6..2ad477c99 100644
--- a/server/tests/api/server/config.ts
+++ b/server/tests/api/server/config.ts
@@ -52,6 +52,7 @@ function checkInitialConfig (server: ServerInfo, data: CustomConfig) {
   expect(data.user.videoQuotaDaily).to.equal(-1)
   expect(data.transcoding.enabled).to.be.false
   expect(data.transcoding.allowAdditionalExtensions).to.be.false
+  expect(data.transcoding.allowAudioFiles).to.be.false
   expect(data.transcoding.threads).to.equal(2)
   expect(data.transcoding.resolutions['240p']).to.be.true
   expect(data.transcoding.resolutions['360p']).to.be.true
@@ -102,6 +103,7 @@ function checkUpdatedConfig (data: CustomConfig) {
   expect(data.transcoding.enabled).to.be.true
   expect(data.transcoding.threads).to.equal(1)
   expect(data.transcoding.allowAdditionalExtensions).to.be.true
+  expect(data.transcoding.allowAudioFiles).to.be.true
   expect(data.transcoding.resolutions['240p']).to.be.false
   expect(data.transcoding.resolutions['360p']).to.be.true
   expect(data.transcoding.resolutions['480p']).to.be.true
@@ -215,6 +217,7 @@ describe('Test config', function () {
       transcoding: {
         enabled: true,
         allowAdditionalExtensions: true,
+        allowAudioFiles: true,
         threads: 1,
         resolutions: {
           '240p': false,