ffmpeg.ts
| 1 | import { AudioCodec, MediaContainer, VideoCodec, VideoEncodingSetting } from "music-server-shared/types"; |
| 2 | import { createRFC6381String } from "music-server-shared/utils"; |
| 3 | import { analyzeWithMediainfo, readStream } from "./utils"; |
| 4 | |
| 5 | function secondsToFfmpeg(seconds: number): string { |
| 6 | // Ensure seconds is non-negative |
| 7 | if (seconds < 0) { |
| 8 | throw new Error("Seconds must be a non-negative number"); |
| 9 | } |
| 10 | |
| 11 | // Extract hours, minutes, seconds, and milliseconds |
| 12 | const hours = Math.floor(seconds / 3600); |
| 13 | const minutes = Math.floor((seconds % 3600) / 60); |
| 14 | const secs = Math.floor(seconds % 60); |
| 15 | |
| 16 | // Calculate the fractional part for milliseconds |
| 17 | const milliseconds = Math.floor((seconds - Math.floor(seconds)) * 1000); |
| 18 | |
| 19 | // Pad numbers with leading zeros |
| 20 | const hoursStr = hours.toString().padStart(2, "0"); |
| 21 | const minutesStr = minutes.toString().padStart(2, "0"); |
| 22 | const secondsStr = secs.toString().padStart(2, "0"); |
| 23 | const millisecondsStr = milliseconds.toString().padStart(3, "0"); |
| 24 | |
| 25 | return `${hoursStr}:${minutesStr}:${secondsStr}.${millisecondsStr}`; |
| 26 | } |
| 27 | |
| 28 | export async function convertWithFFmpeg( |
| 29 | filePath: string, |
| 30 | audioBitrate: number, |
| 31 | videoBitrate: number, |
| 32 | container: MediaContainer, |
| 33 | requestedAudioCodec: AudioCodec, |
| 34 | requestedVideoCodec: VideoCodec, |
| 35 | requestedVideoEncodingSetting: VideoEncodingSetting, |
| 36 | seekTo: number | undefined, |
| 37 | languages: string, |
| 38 | ) { |
| 39 | const mimeType = createRFC6381String(container, requestedVideoCodec, requestedAudioCodec); |
| 40 | |
| 41 | let audioCodec = undefined; |
| 42 | switch (requestedAudioCodec) { |
| 43 | case AudioCodec.aac: |
| 44 | audioCodec = "aac"; |
| 45 | break; |
| 46 | case AudioCodec.mp3: { |
| 47 | audioCodec = "libmp3lame"; |
| 48 | break; |
| 49 | } |
| 50 | case AudioCodec.opus: { |
| 51 | audioCodec = "libopus"; |
| 52 | break; |
| 53 | } |
| 54 | case AudioCodec.vorbis: { |
| 55 | audioCodec = "libvorbis"; |
| 56 | break; |
| 57 | } |
| 58 | default: { |
| 59 | throw new Error(`Unsupported audio codec: ${requestedAudioCodec}`); |
| 60 | } |
| 61 | } |
| 62 | |
| 63 | let maxWidth = undefined; |
| 64 | if (videoBitrate <= 210000) { |
| 65 | maxWidth = 860; |
| 66 | } else if (videoBitrate < 420000) { |
| 67 | maxWidth = 1000; |
| 68 | } else if (videoBitrate < 1500000) { |
| 69 | maxWidth = 1280; |
| 70 | } else if (videoBitrate < 2000000) { |
| 71 | maxWidth = 1920; |
| 72 | } else if (videoBitrate < 10000000) { |
| 73 | maxWidth = 3840; |
| 74 | } |
| 75 | |
| 76 | let containerSettings = undefined; |
| 77 | if (container === MediaContainer.mp4) { |
| 78 | containerSettings = [ |
| 79 | "-avoid_negative_ts", |
| 80 | "make_zero", |
| 81 | "-movflags", |
| 82 | "+frag_keyframe+empty_moov+default_base_moof+faststart", |
| 83 | ]; |
| 84 | } |
| 85 | |
| 86 | const videoEncodingSettings: string[] = []; |
| 87 | let videoCodec = undefined; |
| 88 | let extraVideoSettings = undefined; |
| 89 | //only convert video if transcoding is requesting so |
| 90 | if (videoCodec !== VideoCodec.none) { |
| 91 | extraVideoSettings = [ |
| 92 | //make sure pixel format is browser compatible |
| 93 | "-pix_fmt", |
| 94 | "yuv420p", |
| 95 | //automatically deinterlace video to double fps if video is interlaced |
| 96 | //scale video down based on bitrate |
| 97 | "-vf", |
| 98 | `yadif=deint=1:mode=1${maxWidth ? `, scale='if(gt(iw,${maxWidth}),${maxWidth},iw)':-2` : ""}`, |
| 99 | //one keyframe ~ 5 seconds for 30 fps video |
| 100 | //important because if this is too long, the frontend might remove frames relevant for current playback when discarding old data from the sourcebuffer |
| 101 | "-g", |
| 102 | "150", |
| 103 | ]; |
| 104 | |
| 105 | switch (requestedVideoCodec) { |
| 106 | case VideoCodec.av1: |
| 107 | videoCodec = "libsvtav1"; //TODO: support hardware encoding |
| 108 | switch (requestedVideoEncodingSetting) { |
| 109 | case VideoEncodingSetting.veryfast: |
| 110 | videoEncodingSettings.push("-preset", "8"); |
| 111 | break; |
| 112 | case VideoEncodingSetting.fast: |
| 113 | videoEncodingSettings.concat("-preset", "7"); |
| 114 | break; |
| 115 | case VideoEncodingSetting.balanced: |
| 116 | videoEncodingSettings.concat("-preset", "6"); |
| 117 | break; |
| 118 | case VideoEncodingSetting.quality: |
| 119 | videoEncodingSettings.concat("-preset", "5"); |
| 120 | break; |
| 121 | default: { |
| 122 | throw new Error(`Unsupported video encoding setting: ${requestedVideoEncodingSetting}`); |
| 123 | } |
| 124 | } |
| 125 | break; |
| 126 | case VideoCodec.h264: |
| 127 | videoCodec = "libx264"; //TODO: support hardware encoding |
| 128 | videoEncodingSettings.concat(["-profile", "high"]); |
| 129 | switch (requestedVideoEncodingSetting) { |
| 130 | case VideoEncodingSetting.veryfast: |
| 131 | videoEncodingSettings.push("-preset", "veryfast"); |
| 132 | break; |
| 133 | case VideoEncodingSetting.fast: |
| 134 | videoEncodingSettings.concat("-preset", "fast"); |
| 135 | break; |
| 136 | case VideoEncodingSetting.balanced: |
| 137 | videoEncodingSettings.concat("-preset", "medium"); |
| 138 | break; |
| 139 | case VideoEncodingSetting.quality: |
| 140 | videoEncodingSettings.concat("-preset", "slow"); |
| 141 | break; |
| 142 | default: { |
| 143 | throw new Error(`Unsupported video encoding setting: ${requestedVideoEncodingSetting}`); |
| 144 | } |
| 145 | } |
| 146 | break; |
| 147 | case VideoCodec.h265: |
| 148 | videoCodec = "libx265"; //TODO: support hardware encoding |
| 149 | videoEncodingSettings.concat(["-profile", "main"]); |
| 150 | switch (requestedVideoEncodingSetting) { |
| 151 | case VideoEncodingSetting.veryfast: |
| 152 | videoEncodingSettings.push("-preset", "veryfast"); |
| 153 | break; |
| 154 | case VideoEncodingSetting.fast: |
| 155 | videoEncodingSettings.concat("-preset", "fast"); |
| 156 | break; |
| 157 | case VideoEncodingSetting.balanced: |
| 158 | videoEncodingSettings.concat("-preset", "medium"); |
| 159 | break; |
| 160 | case VideoEncodingSetting.quality: |
| 161 | videoEncodingSettings.concat("-preset", "slow"); |
| 162 | break; |
| 163 | default: { |
| 164 | throw new Error(`Unsupported video encoding setting: ${requestedVideoEncodingSetting}`); |
| 165 | } |
| 166 | } |
| 167 | break; |
| 168 | case VideoCodec.vp9: |
| 169 | videoCodec = "libvpx-vp9"; |
| 170 | switch (requestedVideoEncodingSetting) { |
| 171 | case VideoEncodingSetting.veryfast: |
| 172 | videoEncodingSettings.concat("-deadline", "realtime", "-cpu-used", "5"); |
| 173 | break; |
| 174 | case VideoEncodingSetting.fast: |
| 175 | videoEncodingSettings.concat("-deadline", "realtime", "-cpu-used", "4"); |
| 176 | break; |
| 177 | case VideoEncodingSetting.balanced: |
| 178 | videoEncodingSettings.concat("-deadline", "good", "-cpu-used", "2"); |
| 179 | break; |
| 180 | case VideoEncodingSetting.quality: |
| 181 | videoEncodingSettings.concat("-deadline", "good", "-cpu-used", "2"); |
| 182 | break; |
| 183 | default: { |
| 184 | throw new Error(`Unsupported video encoding setting: ${requestedVideoEncodingSetting}`); |
| 185 | } |
| 186 | } |
| 187 | break; |
| 188 | } |
| 189 | } |
| 190 | const seekParams = seekTo === undefined ? [] : ["-ss", secondsToFfmpeg(seekTo)]; |
| 191 | |
| 192 | const info = await analyzeWithMediainfo(filePath); //analyze video file with mediainfo to get video duration and other metadata |
| 193 | let selectedAudioTrack = 0; //0 as fallback if no language matches and no default is set |
| 194 | const languageArray = languages.split(",").map((s) => s.toLocaleLowerCase()); |
| 195 | if (languageArray.length > 0) { |
| 196 | let index = 0; |
| 197 | for (const track of info.media?.track || []) { |
| 198 | if (track["@type"] === "Audio") { |
| 199 | if (track.Default === "Yes") selectedAudioTrack = index; //use the default track if no language matches |
| 200 | if (track.Language && languageArray.includes(track.Language)) { |
| 201 | selectedAudioTrack = index; |
| 202 | //use the first matching track |
| 203 | break; |
| 204 | } |
| 205 | index++; |
| 206 | } |
| 207 | } |
| 208 | } |
| 209 | |
| 210 | const args = ["ffmpeg"] |
| 211 | .concat(seekParams) //put seek first to do a fast keyframe seek |
| 212 | .concat(["-i", filePath]) |
| 213 | .concat(containerSettings || []) |
| 214 | .concat( |
| 215 | videoCodec |
| 216 | ? ["-c:v", videoCodec, "-b:v", videoBitrate.toString()] //TODO: better bitrate selection |
| 217 | .concat(extraVideoSettings || []) |
| 218 | .concat(videoEncodingSettings) |
| 219 | .concat(["-map", "0:v:0"]) |
| 220 | : ["-vn"], |
| 221 | ) |
| 222 | .concat(videoEncodingSettings) |
| 223 | .concat(["-map", `0:a:${selectedAudioTrack.toString()}?`]) |
| 224 | .concat(["-c:a", audioCodec, "-b:a", audioBitrate.toString(), "-ac", "2"]) |
| 225 | .concat(["-f", container, "-"]); |
| 226 | console.log(args.join(" ")); |
| 227 | const cmd = Bun.spawn(args, { |
| 228 | stdin: "ignore", |
| 229 | stdout: "pipe", |
| 230 | stderr: "pipe", |
| 231 | onExit: async (subprocess: Bun.Subprocess<"ignore", "pipe", "pipe">, exitCode: number | null) => { |
| 232 | if (exitCode !== 0 && exitCode !== null) { |
| 233 | console.error("ffmpeg failed with status code", exitCode); |
| 234 | console.error(new TextDecoder().decode(await readStream(subprocess.stderr as ReadableStream))); |
| 235 | } |
| 236 | }, |
| 237 | }); |
| 238 | return { cmd, mimeType }; |
| 239 | } |
| 240 |