ai-chatbot/lib/video-canvas/smart-edit/render.ts

import { readFile, writeFile } from "node:fs/promises";
import path from "node:path";
import {
  type EditPlan,
  editOutputSize,
  planTimeline,
  compiledSubtitles,
  subtitlesToSrt,
  subtitlesToAss,
  SECOND,
} from "./contract";
import { ffmpeg, mediaSafety } from "./local";
const renderFfmpeg = (args: string[], cwd?: string) => ffmpeg(args, cwd, 600000);
export async function renderEditPlan(
  plan: EditPlan,
  files: Record<string, string>,
  musicFile: string | undefined,
  dir: string,
  progress: (stage: string) => Promise<void> = async () => {},
) {
  const timeline = planTimeline(plan),
    duration = timeline.duration / SECOND;
  const { width, height } = editOutputSize(plan);
  const normalized: string[] = [];
  for (let i = 0; i < timeline.clips.length; i++) {
    await progress(`处理片段 ${i + 1}/${timeline.clips.length}`);
    const clip = timeline.clips[i],
      asset = plan.assets.find((a) => a.id === clip.assetId)!;
    const output = path.join(dir, `clip-${i}.mp4`);
    const length = (clip.sourceOut - clip.sourceIn) / SECOND;
    const inputs = [
      ...mediaSafety,
      "-ss",
      String(clip.sourceIn / SECOND),
      "-t",
      String(length),
      "-i",
      files[clip.assetId],
    ];
    if (!asset.hasAudio) inputs.push("-f", "lavfi", "-i", "anullsrc=r=48000:cl=stereo");
    await renderFfmpeg([
      ...inputs,
      "-map",
      "0:v:0",
      "-map",
      asset.hasAudio ? "0:a:0" : "1:a:0",
      "-vf",
      `setpts=(PTS-STARTPTS)/${clip.speed},scale=${width}:${height}:force_original_aspect_ratio=decrease,pad=${width}:${height}:(ow-iw)/2:(oh-ih)/2,setsar=1,fps=30,format=yuv420p`,
      "-af",
      `asetpts=PTS-STARTPTS,atempo=${clip.speed},aresample=48000,aformat=channel_layouts=stereo,volume=${clip.gain},apad`,
      "-t",
      String(clip.duration / SECOND),
      "-c:v",
      "libx264",
      "-preset",
      "veryfast",
      "-crf",
      "21",
      "-maxrate",
      "6M",
      "-bufsize",
      "12M",
      "-threads",
      "2",
      "-c:a",
      "aac",
      "-ar",
      "48000",
      "-ac",
      "2",
      output,
    ]);
    normalized.push(output);
  }
  await progress("合成时间轴");
  const joined = path.join(dir, "joined.mp4");
  if (!timeline.clips.some((clip) => clip.overlap) || normalized.length === 1) {
    await writeFile(
      path.join(dir, "concat.txt"),
      normalized.map((_, i) => `file 'clip-${i}.mp4'`).join("\n"),
    );
    await renderFfmpeg(
      ["-f", "concat", "-safe", "1", "-i", "concat.txt", "-c", "copy", joined],
      dir,
    );
  } else {
    // Normalize first: all xfade inputs have identical pixel format, timebase and dimensions.
    const filters: string[] = normalized.map(
      (_, i) => `[${i}:v]settb=AVTB,setpts=PTS-STARTPTS[vin${i}]`,
    );
    let video = "vin0",
      audio = "0:a";
    for (let i = 1; i < normalized.length; i++) {
      const clip = timeline.clips[i];
      if (clip.overlap)
        filters.push(
          `[${video}][vin${i}]xfade=transition=${clip.transitionType}:duration=${clip.overlap / SECOND}:offset=${clip.start / SECOND}[v${i}]`,
          `[${audio}][${i}:a]acrossfade=d=${clip.overlap / SECOND}:c1=tri:c2=tri[a${i}]`,
        );
      else filters.push(`[${video}][${audio}][vin${i}][${i}:a]concat=n=2:v=1:a=1[v${i}][a${i}]`);
      video = `v${i}`;
      audio = `a${i}`;
    }
    await renderFfmpeg([
      ...normalized.flatMap((file) => ["-threads", "2", "-i", file]),
      "-filter_complex",
      filters.join(";"),
      "-map",
      `[${video}]`,
      "-map",
      `[${audio}]`,
      "-t",
      String(duration),
      "-c:v",
      "libx264",
      "-preset",
      "veryfast",
      "-crf",
      "21",
      "-maxrate",
      "6M",
      "-bufsize",
      "12M",
      "-threads",
      "2",
      "-c:a",
      "aac",
      joined,
    ]);
  }
  const subtitles = compiledSubtitles(plan);
  const srt = path.join(dir, "subtitles.srt");
  await writeFile(srt, subtitlesToSrt(subtitles));
  await writeFile(path.join(dir, "subtitles.ass"), subtitlesToAss(plan, width, height));
  const output = path.join(dir, "result.mp4");
  const inputs = ["-i", joined];
  const filters: string[] = [];
  if (subtitles.length) filters.push("[0:v]ass=filename=subtitles.ass[v]");
  if (musicFile && plan.music) {
    inputs.push("-ss", String(plan.music.start / SECOND), "-i", musicFile);
    filters.push(
      `[1:a]asetpts=PTS-STARTPTS,aresample=48000,aformat=channel_layouts=stereo,volume=${plan.music.gain},apad,atrim=duration=${duration}[music]`,
      `[0:a][music]amix=inputs=2:duration=first:normalize=0,alimiter=limit=0.95:level=0[a]`,
    );
  }
  await progress("输出成片");
  await renderFfmpeg(
    [
      ...inputs,
      ...(filters.length ? ["-filter_complex", filters.join(";")] : []),
      "-map",
      subtitles.length ? "[v]" : "0:v:0",
      "-map",
      musicFile && plan.music ? "[a]" : "0:a:0",
      "-c:v",
      subtitles.length ? "libx264" : "copy",
      ...(subtitles.length
        ? [
            "-preset",
            "veryfast",
            "-crf",
            "21",
            "-maxrate",
            "6M",
            "-bufsize",
            "12M",
            "-threads",
            "2",
          ]
        : []),
      "-c:a",
      "aac",
      "-t",
      String(duration),
      "-movflags",
      "+faststart",
      output,
    ],
    dir,
  );
  return { file: output, srt, edl: JSON.stringify(plan, null, 2), width, height, duration };
}