jianying-subtitle

Transcribe audio/video to subtitles using the JianYing (CapCut) free ASR API. Works as a CLI, a TypeScript library, and a pi extension.

• Audio formats: mp3, wav, flac, m4a (uploaded directly)
• Video: any format ffmpeg can read (audio track is extracted automatically; supported codecs are stream-copied without re-encoding, otherwise transcoded to WAV; requires ffmpeg/ffprobe on PATH)
• Output: srt, ass, json (with word-level timestamps), txt
• No API key required

Privacy

Audio is uploaded to ByteDance (CapCut) cloud servers for processing. Do not use for confidential content.

CLI

npm install -g jianying-subtitle

jianying-subtitle input.mp4                      # → input.json
jianying-subtitle input.mp3 -o subs/out.srt      # custom output path
jianying-subtitle input.wav -o out.json          # format inferred from extension
jianying-subtitle input.mp4 -v                   # verbose progress

Library

npm install jianying-subtitle

import { transcribe, writeSubtitles, toSrt } from "jianying-subtitle";

const segments = await transcribe({
  input: "audio.mp3",
  onProgress: (percent, message) => console.log(percent, message),
});

console.log(toSrt(segments));
writeSubtitles(segments, "out", "srt", { also: ["json"] });

Subpath exports: jianying-subtitle/asr, jianying-subtitle/subtitle.

pi extension

Install as a pi package:

pi install npm:jianying-subtitle

This registers a transcribe_media tool the model can call, e.g.:

"Transcribe ./talk.mp4 to srt and txt"

The tool extracts video audio if needed (fast lossless stream copy when the codec is already supported), streams progress, waits for ASR to finish, writes the subtitle files, and returns a transcript preview in the same tool result (synchronous — no background job / polling).

How it works

Upload → submit ASR task → poll → parse, against the same endpoints the JianYing desktop client uses. See src/asr/ for details.

License

GPL-3.0-only