Create Chinese subtitles and a Markdown transcript from a permitted video

xiaohu-video-translate is a set of open-source Skills for Claude Code, Codex, Gemini CLI, and similar coding Agents. It connects video download or local-file input, Whisper transcription, translation, subtitle polishing, burn-in, and Markdown export.

What it does

Use a permitted URL or local video. You can stop at a transcript or run the full workflow to download video and existing subtitles, create timestamped text, translate it into Chinese or bilingual subtitles, refine line breaks, burn subtitles into the video, and export searchable Markdown. The upstream separates this work into xiaohu-video-md, subtitle-polish, and video-download, so transcription, translation, and rendering need not be repeated together.

Start safely

1. Confirm the right to download, translate, and republish the video.
2. Test a short clip, then manually correct the source transcript before translating.
3. Keep a terminology list for names, products, and technical terms.
4. Check line length, segmentation, and timing before burn-in.
5. Retain the source video, subtitle files, and Markdown instead of only an irreversible burned-in copy.

The project is tuned primarily for macOS and Apple Silicon, where MLX Whisper downloads an approximately 1.5 GB model on first use. On Windows, WSL2 with Ubuntu is the simpler route. Native Windows needs faster-whisper, Git Bash or manual Skill installation, and a Chinese-capable font such as Microsoft YaHei; Linux needs an installed CJK font such as Noto Sans CJK.

Privacy, platform, and rights limits

Transcription can run locally, but subtitle translation uses your active AI Agent and can still send content to a cloud model. Video downloads may require cookies, login state, or a proxy. Treat configuration and login data as sensitive and never commit them publicly. Background music, accents, overlapping speakers, platform changes, and anti-bot measures can reduce reliability. Tool capability does not grant copyright or platform rights.

Source and license

• Upstream repository: https://github.com/xiaohuailabs/xiaohu-video-translate
• GitHub Releases: https://github.com/xiaohuailabs/xiaohu-video-translate/releases
• Upstream owner: xiaohuailabs
• License: MIT
• Release status: no release or tag found at capture
• Snapshot commit: c7836c6319a5
• Captured: 2026-08-09

GitHub currently provides no packaged Release. WSL2 is the simplest Windows path; native Windows uses faster-whisper and requires a local CJK font such as Microsoft YaHei for subtitle burn-in.