video-subtitle

# Video Subtitle Pipeline / 视频字幕流水线 Complete workflow for adding subtitles to videos: **audio extraction → speech-to-text → translation → subtitle burning** 完整的视频字幕生成流程: **提取音频 → 语音转文字 → 翻译 → 烧录字幕** > Read [`references/style-presets.md`](references/style-presets.md) for the full > ASS style reference. Read [`references/troubleshooting.md`](references/troubleshooting.md) > when something doesn't render correctly. ## Prerequisites / 前置条件 - `ffmpeg` and `ffprobe` in PATH — verify with `ffmpeg -version` - Python 3 with `openai-whisper` — `pip install openai-whisper` - macOS / Linux / Windows - First Whisper run downloads a model (~140MB for `base`) > Run `npx video-subtitle-verify` (or `node bin/verify.js`) at any time to > check your environment. ## 快速开始 / Quick Start Tell your agent: "add English subtitles to lecture.mp4" or "帮我把 lecture.mp4 生成中英双语字幕". The agent will automatically invoke the pipeline. The four steps are described below in case you want to call them manually. ## Workflow ### Step 1 — Extract Audio / 提取音频 ```bash python3 scripts/extract_audio.py "<VIDEO_PATH>" --output "<AUDIO_PATH>.wav" ``` If `--output` is omitted, defaults to `<video_stem>.wav`. Output is 16kHz mono WAV — best for Whisper accuracy. ### Step 2 — Transcribe (Whisper) / 语音转写 ```bash python3 scripts/transcribe.py "<AUDIO_PATH>" --model base --output "<SUBTITLE_PATH>.srt" ``` Models: `tiny` (fastest) · `base` (good default) · `small` · `medium` · `large` (best accuracy). First run d...