@remotion/renderer/dist/combine-audio.js

Render Remotion videos using Node.js or Bun

"use strict";
Object.defineProperty(exports, "__esModule", { value: true });
exports.createCombinedAudio = exports.getClosestAlignedTime = exports.durationOf1Frame = void 0;
const fs_1 = require("fs");
const path_1 = require("path");
const version_1 = require("remotion/version");
const call_ffmpeg_1 = require("./call-ffmpeg");
const logger_1 = require("./logger");
const audio_codec_1 = require("./options/audio-codec");
const parse_ffmpeg_progress_1 = require("./parse-ffmpeg-progress");
const sample_rate_1 = require("./sample-rate");
const truthy_1 = require("./truthy");
exports.durationOf1Frame = (1024 / sample_rate_1.DEFAULT_SAMPLE_RATE) * 1000000;
const roundWithFix = (targetTime) => {
    // Round values where the fractional part is > 0.4999999 up to the next integer,
    // otherwise round down. This addresses floating-point precision issues that can
    // lead to audio imperfections, such as demonstrated in https://github.com/remotion-dev/remotion/issues/6010
    if (targetTime % 1 > 0.4999999) {
        return Math.ceil(targetTime);
    }
    return Math.floor(targetTime);
};
const getClosestAlignedTime = (targetTime) => {
    const decimalFramesToTargetTime = (targetTime * 1000000) / exports.durationOf1Frame;
    const nearestFrameIndexForTargetTime = roundWithFix(decimalFramesToTargetTime);
    return (nearestFrameIndexForTargetTime * exports.durationOf1Frame) / 1000000;
};
exports.getClosestAlignedTime = getClosestAlignedTime;
const encodeAudio = async ({ files, resolvedAudioCodec, audioBitrate, filelistDir, output, indent, logLevel, addRemotionMetadata, fps, binariesDirectory, cancelSignal, onProgress, }) => {
    var _a;
    const fileList = files.map((p) => `file '${p}'`).join('\n');
    const fileListTxt = (0, path_1.join)(filelistDir, 'audio-files.txt');
    (0, fs_1.writeFileSync)(fileListTxt, fileList);
    const startCombining = Date.now();
    const command = [
        '-hide_banner',
        '-f',
        'concat',
        '-safe',
        '0',
        '-i',
        fileListTxt,
        '-c:a',
        (0, audio_codec_1.mapAudioCodecToFfmpegAudioCodecName)(resolvedAudioCodec),
        resolvedAudioCodec === 'aac' ? '-cutoff' : null,
        resolvedAudioCodec === 'aac' ? '18000' : null,
        '-b:a',
        audioBitrate ? audioBitrate : '320k',
        '-vn',
        addRemotionMetadata ? `-metadata` : null,
        addRemotionMetadata ? `comment=Made with Remotion ${version_1.VERSION}` : null,
        '-y',
        output,
    ];
    logger_1.Log.verbose({ indent, logLevel }, `Combining audio with re-encoding, command: ${command.join(' ')}`);
    try {
        const task = (0, call_ffmpeg_1.callFf)({
            args: command,
            bin: 'ffmpeg',
            indent,
            logLevel,
            binariesDirectory,
            cancelSignal,
        });
        (_a = task.stderr) === null || _a === void 0 ? void 0 : _a.on('data', (data) => {
            const utf8 = data.toString('utf8');
            const parsed = (0, parse_ffmpeg_progress_1.parseFfmpegProgress)(utf8, fps);
            if (parsed === undefined) {
                logger_1.Log.verbose({ indent, logLevel }, utf8);
            }
            else {
                onProgress(parsed);
                logger_1.Log.verbose({ indent, logLevel }, `Encoded ${parsed} audio frames`);
            }
        });
        await task;
        logger_1.Log.verbose({ indent, logLevel }, `Encoded audio in ${Date.now() - startCombining}ms`);
        return output;
    }
    catch (e) {
        (0, fs_1.rmSync)(fileListTxt, { recursive: true });
        throw e;
    }
};
const combineAudioSeamlessly = async ({ files, filelistDir, indent, logLevel, output, chunkDurationInSeconds, addRemotionMetadata, fps, binariesDirectory, cancelSignal, onProgress, }) => {
    var _a;
    const startConcatenating = Date.now();
    const fileList = files
        .map((p, i) => {
        const isLast = i === files.length - 1;
        const targetStart = i * chunkDurationInSeconds;
        const endStart = (i + 1) * chunkDurationInSeconds;
        const startTime = (0, exports.getClosestAlignedTime)(targetStart) * 1000000;
        const endTime = (0, exports.getClosestAlignedTime)(endStart) * 1000000;
        const realDuration = endTime - startTime;
        let inpoint = 0;
        if (i > 0) {
            // Although we only asked for two frames of padding, ffmpeg will add an
            // additional 2 frames of silence at the start of the segment. When we slice out
            // our real data with inpoint and outpoint, we'll want remove both the silence
            // and the extra frames we asked for.
            inpoint = exports.durationOf1Frame * 4;
        }
        // inpoint is inclusive and outpoint is exclusive. To avoid overlap, we subtract
        // the duration of one frame from the outpoint.
        // we don't have to subtract a frame if this is the last segment.
        const outpoint = (i === 0 ? exports.durationOf1Frame * 2 : inpoint) +
            realDuration -
            (isLast ? 0 : exports.durationOf1Frame);
        return [`file '${p}'`, `inpoint ${inpoint}us`, `outpoint ${outpoint}us`]
            .filter(truthy_1.truthy)
            .join('\n');
    })
        .join('\n');
    const fileListTxt = (0, path_1.join)(filelistDir, 'audio-files.txt');
    (0, fs_1.writeFileSync)(fileListTxt, fileList);
    const command = [
        '-hide_banner',
        '-f',
        'concat',
        '-safe',
        '0',
        '-i',
        fileListTxt,
        '-c:a',
        'copy',
        '-vn',
        addRemotionMetadata ? `-metadata` : null,
        addRemotionMetadata ? `comment=Made with Remotion ${version_1.VERSION}` : null,
        '-y',
        output,
    ];
    logger_1.Log.verbose({ indent, logLevel }, `Combining AAC audio seamlessly, command: ${command.join(' ')}`);
    try {
        const task = (0, call_ffmpeg_1.callFf)({
            args: command,
            bin: 'ffmpeg',
            indent,
            logLevel,
            binariesDirectory,
            cancelSignal,
        });
        (_a = task.stderr) === null || _a === void 0 ? void 0 : _a.on('data', (data) => {
            const utf8 = data.toString('utf8');
            const parsed = (0, parse_ffmpeg_progress_1.parseFfmpegProgress)(utf8, fps);
            if (parsed !== undefined) {
                onProgress(parsed);
                logger_1.Log.verbose({ indent, logLevel }, `Encoded ${parsed} audio frames`);
            }
        });
        await task;
        logger_1.Log.verbose({ indent, logLevel }, `Combined audio seamlessly in ${Date.now() - startConcatenating}ms`);
        return output;
    }
    catch (e) {
        (0, fs_1.rmSync)(fileListTxt, { recursive: true });
        logger_1.Log.error({ indent, logLevel }, e);
        throw e;
    }
};
const createCombinedAudio = ({ seamless, filelistDir, files, indent, logLevel, audioBitrate, resolvedAudioCodec, output, chunkDurationInSeconds, addRemotionMetadata, binariesDirectory, fps, cancelSignal, onProgress, }) => {
    if (seamless) {
        return combineAudioSeamlessly({
            filelistDir,
            files,
            indent,
            logLevel,
            output,
            chunkDurationInSeconds,
            addRemotionMetadata,
            binariesDirectory,
            fps,
            cancelSignal,
            onProgress,
        });
    }
    return encodeAudio({
        filelistDir,
        files,
        resolvedAudioCodec,
        audioBitrate,
        output,
        indent,
        logLevel,
        addRemotionMetadata,
        binariesDirectory,
        fps,
        cancelSignal,
        onProgress,
    });
};
exports.createCombinedAudio = createCombinedAudio;