FuerDieCuts/server/render.js

import { spawn } from 'node:child_process';
import crypto from 'node:crypto';
import fsp from 'node:fs/promises';
import path from 'node:path';
import { TARGET_SECONDS } from './settings.js';

export const OUTPUT = { width: 1920, height: 1080, fps: 30, sampleRate: 48000 };

// Erhöhen, wenn sich die Filterkette ändert: dann verfallen zwischengespeicherte Segmente.
const RENDER_VERSION = 1;
const MUSIC_FADE_SECONDS = 2;

const AUDIO_FORMAT = `aformat=sample_fmts=fltp:sample_rates=${OUTPUT.sampleRate}:channel_layouts=stereo`;
// Zwischenstände tragen unkomprimierten Ton; erst das Endergebnis bekommt AAC.
// So summieren sich über viele Übergänge keine Verschiebungen auf.
const SEGMENT_CODECS = [
  '-c:v', 'libx264', '-preset', 'veryfast', '-crf', '20', '-pix_fmt', 'yuv420p',
  '-r', String(OUTPUT.fps), '-video_track_timescale', '15360',
  '-c:a', 'pcm_s16le',
];

const hash = (value) =>
  crypto.createHash('sha256').update(JSON.stringify(value)).digest('hex').slice(0, 32);
const factor = (volume) => Math.round(volume) / 100;
const seconds = (value) => value.toFixed(3);

function readyAsset(project, slot) {
  const asset = project.assets?.[slot];
  return asset?.status === 'ready' ? asset : null;
}

// Beschreibt, was gerendert wird – ohne Pfade, damit der Plan vergleichbar bleibt.
export function planRender(project) {
  const { settings } = project;

  const bookend = (slot) => {
    const asset = readyAsset(project, slot);
    if (!settings[slot].enabled || !asset) return null;
    return {
      kind: slot,
      target: { type: 'asset', slot },
      source: `${slot}:${asset.revision}`,
      speed: 1,
      outDuration: asset.media.duration,
      audio: asset.media.hasAudio ? { volume: factor(settings[slot].volume), tempo: 1 } : null,
      voice: null,
    };
  };

  const clips = project.clips
    .filter((clip) => clip.status === 'ready' && !settings.excluded.includes(clip.index))
    .map((clip) => {
      const { duration, hasAudio } = clip.media;
      const speed = settings.speedUp && duration > TARGET_SECONDS ? duration / TARGET_SECONDS : 1;
      const voice = readyAsset(project, `voice-${clip.index}`);
      return {
        kind: 'clip',
        index: clip.index,
        target: { type: 'clip', index: clip.index },
        source: `clip:${clip.index}:${clip.file}:${clip.size}`,
        speed,
        outDuration: duration / speed,
        audio: !settings.muteClips && hasAudio ? { volume: factor(settings.clipVolume), tempo: speed } : null,
        voice: voice && {
          target: { type: 'asset', slot: `voice-${clip.index}` },
          source: `voice:${voice.revision}`,
          volume: factor(settings.voiceVolume) * factor(settings.voiceVolumes[clip.index] ?? 100),
        },
      };
    });

  const intro = bookend('intro');
  const outro = bookend('outro');
  const segments = [intro, ...clips, outro].filter(Boolean);
  for (const segment of segments) {
    const { source, speed, outDuration, audio, voice } = segment;
    segment.key = hash({
      version: RENDER_VERSION,
      source,
      speed,
      outDuration,
      audio,
      voice: voice && { source: voice.source, volume: voice.volume },
    });
  }

  const clipsDuration = clips.reduce((sum, clip) => sum + clip.outDuration, 0);
  const musicAsset = readyAsset(project, 'music');
  const music =
    settings.music.enabled && musicAsset && clips.length > 0
      ? {
          target: { type: 'asset', slot: 'music' },
          source: `music:${musicAsset.revision}`,
          volume: factor(settings.music.volume),
          duration: clipsDuration,
        }
      : null;

  return {
    intro,
    clips,
    outro,
    music,
    clipsDuration,
    duration: segments.reduce((sum, segment) => sum + segment.outDuration, 0),
    key: hash({
      segments: segments.map((segment) => segment.key),
      music: music && { source: music.source, volume: music.volume },
    }),
  };
}

// atempo schafft je Stufe höchstens Faktor 2, also verketten.
function tempoFilters(tempo) {
  const filters = [];
  let rest = tempo;
  while (rest > 2) {
    filters.push('atempo=2');
    rest /= 2;
  }
  if (rest !== 1) filters.push(`atempo=${rest.toFixed(6)}`);
  return filters;
}

// Ein Segment: auf 1920×1080 bei 30 fps, Hochkant mit unscharfem Hintergrund,
// Ton aus Clip (oder Stille) plus Voice-over, exakt auf Segmentlänge.
export function segmentArgs(segment, inputs, output) {
  const { width, height, fps } = OUTPUT;
  const duration = seconds(segment.outDuration);
  const speed = segment.speed === 1 ? '' : `setpts=PTS/${segment.speed.toFixed(6)},`;

  const filters = [
    `[0:v]scale=iw*sar:ih,setsar=1,${speed}fps=${fps},split=2[bgsrc][fgsrc]`,
    `[bgsrc]scale=${width / 4}:${height / 4}:force_original_aspect_ratio=increase,` +
      `crop=${width / 4}:${height / 4},boxblur=8:2,scale=${width}:${height}[bg]`,
    `[fgsrc]scale=${width}:${height}:force_original_aspect_ratio=decrease[fg]`,
    '[bg][fg]overlay=(W-w)/2:(H-h)/2,setsar=1,format=yuv420p[v]',
    segment.audio
      ? `[0:a]${[...tempoFilters(segment.audio.tempo), `volume=${segment.audio.volume}`, AUDIO_FORMAT].join(',')}[base]`
      : `anullsrc=r=${OUTPUT.sampleRate}:cl=stereo,${AUDIO_FORMAT}[base]`,
  ];
  if (segment.voice) {
    filters.push(`[1:a]volume=${segment.voice.volume},${AUDIO_FORMAT}[voice]`);
    filters.push('[base][voice]amix=inputs=2:duration=first:normalize=0[mixed]');
  }
  filters.push(`[${segment.voice ? 'mixed' : 'base'}]apad,atrim=0:${duration}[a]`);

  return [
    '-i', inputs.video,
    ...(segment.voice ? ['-i', inputs.voice] : []),
    '-filter_complex', filters.join(';'),
    '-map', '[v]', '-map', '[a]', '-t', duration,
    ...SEGMENT_CODECS,
    output,
  ];
}

// Musik unter den Clip-Teil: wiederholt bis zum Ende, dort gekürzt und ausgeblendet.
export function musicArgs(music, inputs, output) {
  const duration = seconds(music.duration);
  const fade = Math.min(MUSIC_FADE_SECONDS, music.duration);
  const fadeStart = seconds(Math.max(0, music.duration - fade));
  return [
    '-i', inputs.middle,
    '-stream_loop', '-1', '-i', inputs.music,
    '-filter_complex',
    `[1:a]${AUDIO_FORMAT},volume=${music.volume},atrim=0:${duration},` +
      `afade=t=out:st=${fadeStart}:d=${seconds(fade)}[m];` +
      '[0:a][m]amix=inputs=2:duration=first:normalize=0[a]',
    '-map', '0:v', '-map', '[a]', '-c:v', 'copy', '-c:a', 'pcm_s16le',
    output,
  ];
}

export function concatArgs(listFile, output, { final }) {
  return [
    '-f', 'concat', '-safe', '0', '-i', listFile,
    '-map', '0:v', '-map', '0:a', '-c:v', 'copy',
    ...(final ? ['-c:a', 'aac', '-b:a', '192k', '-movflags', '+faststart'] : ['-c:a', 'copy']),
    output,
  ];
}

// Startet ffmpeg und meldet die bereits erzeugte Ausgabezeit in Sekunden.
export function runFfmpeg(args, { onTime, signal } = {}) {
  return new Promise((resolve, reject) => {
    if (signal?.aborted) return reject(new Error('Abgebrochen.'));
    const child = spawn(
      'ffmpeg',
      ['-hide_banner', '-nostdin', '-y', '-v', 'error', '-progress', 'pipe:1', ...args],
      { stdio: ['ignore', 'pipe', 'pipe'] },
    );
    let stderr = '';
    let buffer = '';
    const abort = () => child.kill('SIGKILL');
    signal?.addEventListener('abort', abort, { once: true });

    child.stdout.on('data', (data) => {
      buffer += data;
      const lines = buffer.split('\n');
      buffer = lines.pop();
      for (const line of lines) {
        const match = /^out_time_us=(\d+)/.exec(line);
        if (match && onTime) onTime(Number(match[1]) / 1e6);
      }
    });
    child.stderr.on('data', (data) => {
      stderr = (stderr + data).slice(-4000);
    });
    child.on('error', reject);
    child.on('close', (code) => {
      signal?.removeEventListener('abort', abort);
      if (signal?.aborted) reject(new Error('Abgebrochen.'));
      else if (code === 0) resolve();
      else reject(new Error(`ffmpeg ist fehlgeschlagen: ${stderr.trim().split('\n').slice(-3).join(' | ')}`));
    });
  });
}

async function exists(file) {
  try {
    await fsp.access(file);
    return true;
  } catch {
    return false;
  }
}

async function writeConcatList(file, entries) {
  const quote = (entry) => `file '${entry.replace(/'/g, "'\\''")}'`;
  await fsp.writeFile(file, `${entries.map(quote).join('\n')}\n`);
}

// Rendert den Plan nach output. resolve(target) liefert den Pfad einer Eingabedatei.
export async function renderPlan({ plan, resolve, workDir, output, onProgress = () => {}, signal }) {
  if (plan.clips.length === 0) throw new Error('Keine Clips für das Video ausgewählt.');

  const cacheDir = path.join(workDir, 'cache');
  const tempDir = path.join(workDir, 'tmp');
  await fsp.mkdir(cacheDir, { recursive: true });
  await fsp.rm(tempDir, { recursive: true, force: true });
  await fsp.mkdir(tempDir, { recursive: true });

  // Gewichte für die Fortschrittsanzeige: Segmente kosten fast die ganze Zeit.
  const musicWeight = plan.music ? plan.clipsDuration * 0.1 : 0;
  const finalWeight = plan.duration * 0.05;
  const total = plan.duration + musicWeight + finalWeight;
  let done = 0;
  const report = (current = 0) => onProgress(Math.min(1, (done + current) / total));

  const segmentFile = (segment) => path.join(cacheDir, `${segment.key}.mov`);
  const segments = [plan.intro, ...plan.clips, plan.outro].filter(Boolean);
  for (const segment of segments) {
    const file = segmentFile(segment);
    if (!(await exists(file))) {
      const temp = path.join(tempDir, `${segment.key}.mov`);
      const inputs = {
        video: resolve(segment.target),
        voice: segment.voice ? resolve(segment.voice.target) : null,
      };
      await runFfmpeg(segmentArgs(segment, inputs, temp), {
        signal,
        onTime: (time) => report(Math.min(time, segment.outDuration)),
      });
      await fsp.rename(temp, file);
    }
    done += segment.outDuration;
    report();
  }

  const middleList = path.join(tempDir, 'middle.txt');
  let middle = path.join(tempDir, 'middle.mov');
  await writeConcatList(middleList, plan.clips.map(segmentFile));
  await runFfmpeg(concatArgs(middleList, middle, { final: false }), { signal });

  if (plan.music) {
    const withMusic = path.join(tempDir, 'middle-music.mov');
    await runFfmpeg(musicArgs(plan.music, { middle, music: resolve(plan.music.target) }, withMusic), {
      signal,
      onTime: (time) => report((Math.min(time, plan.clipsDuration) / plan.clipsDuration) * musicWeight),
    });
    middle = withMusic;
    done += musicWeight;
    report();
  }

  const finalList = path.join(tempDir, 'final.txt');
  const parts = [plan.intro && segmentFile(plan.intro), middle, plan.outro && segmentFile(plan.outro)];
  await writeConcatList(finalList, parts.filter(Boolean));
  const tempOutput = path.join(tempDir, 'output.mp4');
  await runFfmpeg(concatArgs(finalList, tempOutput, { final: true }), {
    signal,
    onTime: (time) => report((Math.min(time, plan.duration) / plan.duration) * finalWeight),
  });
  await fsp.rename(tempOutput, output);
  onProgress(1);

  // Aufräumen: Zwischenstände und Segmente, die der aktuelle Plan nicht mehr braucht.
  await fsp.rm(tempDir, { recursive: true, force: true });
  const keep = new Set(segments.map((segment) => `${segment.key}.mov`));
  for (const entry of await fsp.readdir(cacheDir)) {
    if (!keep.has(entry)) await fsp.rm(path.join(cacheDir, entry), { force: true });
  }
}