skill-mv/lib/ffmpeg.mjs

// lib/ffmpeg.mjs — ffmpeg / ffprobe 封装
//
// 只用系统 ffmpeg（/usr/bin/ffmpeg）。零 npm 依赖。
// 所有对外部命令的调用都走 run()，统一超时、统一错误信息（带 stderr 尾部）。

import { spawn, spawnSync } from 'node:child_process';
import { mkdirSync, existsSync, writeFileSync, readdirSync, statSync } from 'node:fs';
import { dirname, join, resolve, basename } from 'node:path';
import { safeEnv } from './config.mjs';

export const FFMPEG = process.env.FFMPEG_BIN || 'ffmpeg';
export const FFPROBE = process.env.FFPROBE_BIN || 'ffprobe';

/** 检查 ffmpeg / ffprobe 是否可用 */
export function checkFfmpeg() {
  const out = {};
  for (const [name, bin] of [
    ['ffmpeg', FFMPEG],
    ['ffprobe', FFPROBE],
  ]) {
    const r = spawnSync(bin, ['-version'], { encoding: 'utf8', timeout: 15_000 });
    out[name] = r.status === 0
      ? { available: true, version: (r.stdout || '').split('\n')[0].trim() }
      : { available: false, version: null, error: (r.stderr || '').split('\n')[0] };
  }
  return out;
}

export function requireFfmpeg() {
  const c = checkFfmpeg();
  if (!c.ffmpeg.available || !c.ffprobe.available) {
    throw new Error(
      '缺少系统依赖 ffmpeg / ffprobe。\n' +
        '  Ubuntu/Debian: sudo apt-get install -y ffmpeg\n' +
        '  macOS: brew install ffmpeg\n' +
        '（当前 ffmpeg: ' + (c.ffmpeg.version || '未找到') + '）'
    );
  }
  return c;
}

/**
 * 执行 ffmpeg（或任意命令）。
 * @param {string[]} args
 * @param {{bin?:string, timeout?:number, label?:string, cwd?:string}} opts
 */
export function run(args, opts = {}) {
  const { bin = FFMPEG, timeout = 30 * 60_000, label = 'ffmpeg', cwd } = opts;
  return new Promise((resolvePromise, reject) => {
    const p = spawn(bin, args, { cwd, env: safeEnv() });
    let stderr = '';
    let stdout = '';
    const timer = setTimeout(() => {
      p.kill('SIGKILL');
      reject(new Error(`${label} 超时（${Math.round(timeout / 1000)}s）`));
    }, timeout);
    p.stdout.on('data', (d) => (stdout += d));
    p.stderr.on('data', (d) => {
      stderr += d;
      if (stderr.length > 200_000) stderr = stderr.slice(-100_000);
    });
    p.on('error', (e) => {
      clearTimeout(timer);
      reject(new Error(`${label} 启动失败：${e.message}（未安装 ${bin}？）`));
    });
    p.on('close', (code) => {
      clearTimeout(timer);
      if (code === 0) return resolvePromise({ stdout, stderr });
      const tail = stderr.split('\n').filter(Boolean).slice(-8).join('\n  ');
      reject(new Error(`${label} 退出码 ${code}：\n  ${tail}`));
    });
  });
}

/** ffprobe 读取媒体信息 */
export async function probe(file) {
  const { stdout } = await run(
    ['-v', 'error', '-print_format', 'json', '-show_format', '-show_streams', file],
    { bin: FFPROBE, label: 'ffprobe', timeout: 60_000 }
  );
  const j = JSON.parse(stdout);
  const v = (j.streams || []).find((s) => s.codec_type === 'video');
  const a = (j.streams || []).find((s) => s.codec_type === 'audio');
  const num = (x) => (x == null || x === 'N/A' ? null : Number(x));
  return {
    raw: j,
    duration: num(j.format?.duration) ?? num(v?.duration) ?? 0,
    size: num(j.format?.size) ?? (existsSync(file) ? statSync(file).size : 0),
    hasVideo: !!v,
    hasAudio: !!a,
    width: num(v?.width),
    height: num(v?.height),
    fps: parseFps(v?.r_frame_rate),
    vcodec: v?.codec_name || null,
    acodec: a?.codec_name || null,
    sampleRate: num(a?.sample_rate),
    channels: num(a?.channels),
  };
}

function parseFps(r) {
  if (!r || r === 'N/A') return null;
  const [a, b] = String(r).split('/').map(Number);
  if (!b) return a || null;
  return +(a / b).toFixed(3);
}

export function ensureDir(dir) {
  mkdirSync(dir, { recursive: true });
  return dir;
}

export function fileSize(p) {
  try {
    return statSync(p).size;
  } catch {
    return 0;
  }
}

export function humanSize(bytes) {
  if (!bytes) return '0B';
  const u = ['B', 'KB', 'MB', 'GB'];
  let i = 0;
  let n = bytes;
  while (n >= 1024 && i < u.length - 1) {
    n /= 1024;
    i++;
  }
  return `${n.toFixed(n < 10 && i > 0 ? 1 : 0)}${u[i]}`;
}

// ==================== 音视频常用操作 ====================

/** 从视频中提取音频 */
export async function extractAudio(input, output, { format = 'wav', sampleRate = 44100, channels = 2 } = {}) {
  ensureDir(dirname(output));
  await run([
    '-y', '-i', input,
    '-vn',
    '-acodec', format === 'wav' ? 'pcm_s16le' : format === 'mp3' ? 'libmp3lame' : 'aac',
    '-ar', String(sampleRate), '-ac', String(channels),
    output,
  ], { label: 'extractAudio' });
  return output;
}

/** 把音频合到视频上（替换音轨），支持淡出与时长裁剪 */
export async function muxAudio(video, audio, output, { duration, fadeOut = 0, audioBitrate = '192k', volume = 1 } = {}) {
  ensureDir(dirname(output));
  const filters = [];
  if (volume !== 1) filters.push(`volume=${volume}`);
  if (fadeOut > 0 && duration) {
    filters.push(`afade=t=out:st=${Math.max(0, duration - fadeOut).toFixed(3)}:d=${fadeOut}`);
  }
  const af = filters.length ? ['-af', filters.join(',')] : [];
  const args = [
    '-y', '-i', video, '-i', audio,
    '-map', '0:v:0', '-map', '1:a:0',
    '-c:v', 'copy',
    '-c:a', 'aac', '-b:a', audioBitrate,
    ...af,
    ...(duration ? ['-t', String(duration)] : []),
    '-shortest',
    output,
  ];
  await run(args, { label: 'muxAudio' });
  return output;
}

/** 拼接多段视频（统一重编码，避免参数不一致导致 concat 失败） */
export async function concat(clips, output, { width = 1280, height = 720, fps = 30, transition = 0, withAudio = false } = {}) {
  ensureDir(dirname(output));
  if (!clips.length) throw new Error('concat 需要至少一个片段');
  if (clips.length === 1) {
    await run(['-y', '-i', clips[0], '-c', 'copy', output], { label: 'concat(single)' });
    return output;
  }

  const args = ['-y'];
  for (const c of clips) args.push('-i', c);

  const scale = `scale=${width}:${height}:force_original_aspect_ratio=decrease,pad=${width}:${height}:(ow-iw)/2:(oh-ih)/2:color=black,setsar=1,fps=${fps}`;

  if (transition > 0) {
    // xfade 交叉溶解
    let filter = '';
    let prev = '[0:v]';
    for (let i = 1; i < clips.length; i++) {
      const off = i * (transition > 0 ? 0.5 : 0);
      const out = i === clips.length - 1 ? '[vout]' : `[v${i}]`;
      filter += `${prev}[${i}:v]xfade=transition=fade:duration=${transition}:offset=${off}${out};`;
      prev = out;
    }
    const mapArgs = ['-map', '[vout]', '-c:v', 'libx264', '-pix_fmt', 'yuv420p', '-crf', '18'];
    await run([...args, '-filter_complex', filter.replace(/;$/, ''), ...mapArgs, output], { label: 'concat(xfade)' });
    return output;
  }

  const fc = clips.map((_, i) => `[${i}:v]${scale}[v${i}]`).join(';') + ';' + clips.map((_, i) => `[v${i}]`).join('') + `concat=n=${clips.length}:v=1:a=0[vout]`;
  await run(
    [...args, '-filter_complex', fc, '-map', '[vout]', '-c:v', 'libx264', '-pix_fmt', 'yuv420p', '-crf', '18', output],
    { label: 'concat' }
  );
  return output;
}

/** 给视频烧硬字幕（SRT / ASS） */
export async function burnSubtitles(video, subtitleFile, output, { fontSize = 20, marginV = 28, fontName = 'Noto Sans CJK SC' } = {}) {
  ensureDir(dirname(output));
  const isAss = subtitleFile.toLowerCase().endsWith('.ass');
  const style = `FontName=${fontName},FontSize=${fontSize},PrimaryColour=&H00FFFFFF,OutlineColour=&H00000000,BorderStyle=1,Outline=2,Shadow=0,Alignment=2,MarginV=${marginV}`;
  const esc = subtitleFile.replace(/\\/g, '/').replace(/:/g, '\\:').replace(/'/g, "\\'");
  const vf = isAss ? `ass='${esc}'` : `subtitles='${esc}':force_style='${style}'`;
  await run(['-y', '-i', video, '-vf', vf, '-c:v', 'libx264', '-pix_fmt', 'yuv420p', '-crf', '18', '-c:a', 'copy', output], {
    label: 'burnSubtitles',
  });
  return output;
}

/** 静帧 + Ken Burns 运镜（视频通道未开通时的兜底动态镜头） */
export async function kenBurns(image, output, { duration = 5, width = 1280, height = 720, fps = 30, zoom = 0.12, direction = 'in' } = {}) {
  ensureDir(dirname(output));
  const frames = Math.max(1, Math.round(duration * fps));
  const zEnd = direction === 'out' ? 1 + zoom : 1 + zoom;
  const zStart = direction === 'out' ? 1 + zoom : 1;
  // zoompan 在 4K 画布上运算以获得平滑慢推
  const pre = `scale=${width * 2}:${height * 2}:force_original_aspect_ratio=increase,crop=${width * 2}:${height * 2}`;
  const zp =
    `zoompan=z='${zStart}+(${zEnd}-${zStart})*on/${frames}':` +
    `x='iw/2-(iw/zoom/2)':y='ih/2-(ih/zoom/2)':d=${frames}:s=${width}x${height}:fps=${fps}`;
  await run(
    ['-y', '-loop', '1', '-i', image, '-t', String(duration), '-vf', `${pre},${zp},format=yuv420p`, '-r', String(fps), '-c:v', 'libx264', '-crf', '19', output],
    { label: 'kenBurns' }
  );
  return output;
}

/** 纯色背景 + 音频（无画面素材时的占位成片） */
export async function colorWithAudio(audio, output, { duration, width = 1280, height = 720, color = 'black', fps = 30 } = {}) {
  ensureDir(dirname(output));
  await run(
    [
      '-y', '-f', 'lavfi', '-i', `color=c=${color}:s=${width}x${height}:r=${fps}`,
      '-i', audio,
      '-map', '0:v', '-map', '1:a',
      ...(duration ? ['-t', String(duration)] : []),
      '-c:v', 'libx264', '-pix_fmt', 'yuv420p', '-c:a', 'aac', '-shortest', output,
    ],
    { label: 'colorWithAudio' }
  );
  return output;
}

/** 生成静音音轨 */
export async function silence(output, duration, { sampleRate = 44100, channels = 2 } = {}) {
  ensureDir(dirname(output));
  await run(
    ['-y', '-f', 'lavfi', '-i', `anullsrc=r=${sampleRate}:cl=${channels === 1 ? 'mono' : 'stereo'}`, '-t', String(duration), '-c:a', 'pcm_s16le', output],
    { label: 'silence' }
  );
  return output;
}

/** 响度归一化（成片听感统一） */
export async function normalizeLoudness(input, output, { target = -14 } = {}) {
  ensureDir(dirname(output));
  const pass1 = await run(['-y', '-i', input, '-af', `loudnorm=I=${target}:print_format=json`, '-f', 'null', '-'], {
    label: 'loudnorm(pass1)',
  });
  let measured = {};
  try {
    const m = pass1.stderr.match(/\{[\s\S]*?"target_offset"[\s\S]*?\}/);
    if (m) measured = JSON.parse(m[0]);
  } catch {
    /* 忽略解析失败，退回单遍 */
  }
  const af = measured.input_i
    ? `loudnorm=I=${target}:TP=-1.5:LRA=11:measured_I=${measured.input_i}:measured_TP=${measured.input_tp}:measured_LRA=${measured.input_lra}:measured_thresh=${measured.input_thresh}:offset=${measured.target_offset}:linear=true`
    : `loudnorm=I=${target}:TP=-1.5:LRA=11`;
  await run(['-y', '-i', input, '-af', af, '-c:a', 'aac', '-b:a', '192k', output], { label: 'loudnorm(pass2)' });
  return output;
}

/** 生成 MP3 伴奏/人声的简易分离（纯 ffmpeg 兜底，见 scripts/separate.mjs 说明） */
export async function separateByChannelPhase(input, { outDir, mode = 'instrumental' } = {}) {
  ensureDir(outDir);
  const out = join(outDir, `${mode}.wav`);
  // 中置声道消除：人声通常在中央，L-R 相减可压低人声 → 近似伴奏
  // 中置提取： (L+R)/2 - (L-R)/2 的近似用 pan 实现
  const af =
    mode === 'instrumental'
      ? 'pan=stereo|c0=c0|c1=-c1,volume=1.4' // 侧信号（近似伴奏）
      : 'pan=mono|c0=0.5*c0+0.5*c1'; // 中置（近似人声/主唱）
  await run(['-y', '-i', input, '-af', af, '-ar', '44100', out], { label: 'separate' });
  return out;
}

/** 用 demucs（若安装）做高质量分离 */
export function checkDemucs() {
  for (const cmd of ['demucs', 'python3']) {
    const r = spawnSync(cmd, cmd === 'python3' ? ['-c', 'import demucs;print(demucs.__version__)'] : ['--help'], {
      encoding: 'utf8',
      timeout: 30_000,
    });
    if (r.status === 0) {
      if (cmd === 'demucs') return { available: true, cmd: 'demucs', version: (r.stdout || '').split('\n')[0] };
      return { available: true, cmd: 'python3 -m demucs', version: (r.stdout || '').trim() };
    }
  }
  return { available: false };
}

/** 用 demucs 分离（可选增强路径） */
export async function separateWithDemucs(input, outDir, { model = 'htdemucs' } = {}) {
  ensureDir(outDir);
  const d = checkDemucs();
  if (!d.available) throw new Error('未检测到 demucs（可选依赖，pip install demucs）');
  const args = d.cmd === 'demucs' ? ['-n', model, '-o', outDir, input] : ['-m', 'demucs', '-n', model, '-o', outDir, input];
  const [bin, ...rest] = d.cmd.split(' ');
  await run([...rest, ...args], { bin, label: 'demucs', timeout: 30 * 60_000 });
  const base = basename(input).replace(/\.[^.]+$/, '');
  return {
    vocals: join(outDir, model, `${base}.vocals.wav`),
    instrumental: join(outDir, model, `${base}.no_vocals.wav`),
    drums: join(outDir, model, `${base}.drums.wav`),
    bass: join(outDir, model, `${base}.bass.wav`),
    other: join(outDir, model, `${base}.other.wav`),
  };
}

// ==================== 卡点：能量包络节拍检测（纯 Node，零依赖）====================

/**
 * 解码音频为 PCM 并做能量包络节拍检测，返回节拍时间点（秒）。
 * 不依赖 aubio 等外部库，够用且稳定。
 * @returns {Promise<{bpm:number|null, beats:number[], duration:number}>}
 */
export async function detectBeats(audioFile, { sampleRate = 11025, hop = 1024, sensitivity = 1.4 } = {}) {
  const tmp = join('/tmp', `fmode-mv-beat-${process.pid}-${Date.now()}.raw`);
  await run(
    ['-y', '-i', audioFile, '-ac', '1', '-ar', String(sampleRate), '-f', 's16le', '-acodec', 'pcm_s16le', tmp],
    { label: 'decode(for beat detection)' }
  );
  const { readFileSync, unlinkSync } = await import('node:fs');
  const buf = readFileSync(tmp);
  try { unlinkSync(tmp); } catch { /* ignore */ }

  const samples = new Int16Array(buf.buffer, buf.byteOffset, Math.floor(buf.length / 2));
  const frames = Math.floor(samples.length / hop);
  const energy = new Float64Array(frames);
  for (let f = 0; f < frames; f++) {
    let sum = 0;
    const off = f * hop;
    for (let i = 0; i < hop; i++) {
      const v = samples[off + i] / 32768;
      sum += v * v;
    }
    energy[f] = Math.sqrt(sum / hop);
  }

  // 局部均值 + 阈值攻陷 → onset 峰值
  const win = 16;
  const beats = [];
  for (let f = 1; f < frames - 1; f++) {
    let sum = 0;
    let n = 0;
    for (let k = Math.max(0, f - win); k <= Math.min(frames - 1, f + win); k++) {
      sum += energy[k];
      n++;
    }
    const local = sum / n;
    const th = local * sensitivity;
    if (energy[f] > th && energy[f] >= energy[f - 1] && energy[f] > energy[f + 1]) {
      const t = (f * hop) / sampleRate;
      if (!beats.length || t - beats[beats.length - 1] > 0.18) beats.push(+t.toFixed(3));
    }
  }

  // 由中位间隔估计 BPM
  let bpm = null;
  if (beats.length > 4) {
    const gaps = [];
    for (let i = 1; i < beats.length; i++) gaps.push(beats[i] - beats[i - 1]);
    gaps.sort((a, b) => a - b);
    const med = gaps[Math.floor(gaps.length / 2)];
    if (med > 0.2) {
      let b = 60 / med;
      while (b < 70) b *= 2;
      while (b > 180) b /= 2;
      bpm = +b.toFixed(1);
    }
  }
  const duration = samples.length / sampleRate;
  return { bpm, beats, duration: +duration.toFixed(3) };
}

/**
 * 把分镜按节拍对齐：给定每镜时长，吸附到最近的节拍点。
 * @param {{beats:number[], duration:number}} beatInfo
 * @param {Array<{duration:number}>} shots
 */
export function alignToBeats(beatInfo, shots, { offset = 0 } = {}) {
  if (!beatInfo?.beats?.length) return shots;
  const out = [];
  let t = offset;
  for (const s of shots) {
    let target = t + s.duration;
    // 找最近的节拍点
    let best = target;
    let bestD = Infinity;
    for (const b of beatInfo.beats) {
      const d = Math.abs(b - target);
      if (d < bestD) {
        bestD = d;
        best = b;
      }
    }
    const snapped = bestD <= 0.6 ? best : target;
    const dur = Math.max(0.8, +(snapped - t).toFixed(3));
    out.push({ ...s, start: +t.toFixed(3), duration: dur });
    t = +(t + dur).toFixed(3);
  }
  return out;
}

/** 列出目录下的媒体文件 */
export function listMedia(dir, exts = ['.mp4', '.mov', '.mkv', '.webm']) {
  if (!existsSync(dir)) return [];
  return readdirSync(dir)
    .filter((f) => exts.includes(f.slice(f.lastIndexOf('.')).toLowerCase()))
    .sort()
    .map((f) => join(dir, f));
}