omni-captions-skills

omni-captions-skills

Captions Made Easy — Claude Code Caption Skills

"I need bilingual captions for this Fireship vibe coding video https://youtube.com/watch?v=Tw18-4U7mts"

One sentence. Claude handles the download, transcription, and translation.

中文文档

Install

npx skills add https://github.com/lattifai/omni-captions-skills

<details> <summary>Other installation methods</summary>

Claude Code Plugin System:

/plugin marketplace add lattifai/omni-captions-skills
/plugin install omnicaptions@lattifai-omni-captions-skills

Local Development:

git clone https://github.com/lattifai/omni-captions-skills.git
claude --plugin-dir ./omni-captions-skills

</details>

Try It

❯ Make bilingual captions for this Fireship vibe coding video https://youtube.com/watch?v=Tw18-4U7mts

1
00:00:00,000 --> 00:00:03,200
Mass hysteria satisfies a deep human need.
群体性癔症满足了人类某种深层需求。

2
00:00:03,200 --> 00:00:07,440
Vibe coding is programming without actually writing any code yourself.
Vibe coding 就是不用自己写代码的编程方式。

Skills

Skill	Description
transcribe	YouTube/video → Markdown with timestamps
translate	Translate captions, bilingual output supported
convert	Convert between 30+ caption formats
download	Download YouTube video/audio/captions
LaiCut	Forced alignment, word-level timing accuracy

Invoke via /omnicaptions:transcribe or /omnicaptions-transcribe

LaiCut: Precise Timing

Standard transcription gives "approximate" timestamps. LaiCut uses LattifAI Lattice-1 model to match text precisely to audio waveforms, achieving word-level accuracy.

Install LaiCut:

# Using uv (recommended, auto-configures package index)
uv pip install "omni-captions-skills[laicut]" --extra-index-url https://lattifai.github.io/pypi/simple/

# Using pip
pip install "omni-captions-skills[laicut]" --extra-index-url https://lattifai.github.io/pypi/simple/

Supported languages: English, Chinese, German, and mixed

Recommended workflow: Align before translate (translated text doesn't match original audio)

Setup

Feature	API Key	Note
Translation	None required	Uses Claude by default, works out of the box
Transcription	Gemini API	Optional, only needed for transcription
LaiCut alignment	LattifAI API	Optional, only needed for precise alignment

Gemini is only used for video transcription. When a video has no captions, you'll be prompted whether to transcribe — configure then. Translation uses Claude by default, works out of the box.

API keys are prompted automatically and saved to ~/.config/omnicaptions/config.json

OmniCaptions CLI Usage

# With captions: download → align → translate
omnicaptions download "https://youtube.com/watch?v=xxx"
omnicaptions LaiCut video.mp4 video.en.vtt -o video_LaiCut.srt
omnicaptions translate video_LaiCut.srt -l zh --bilingual

# Without captions: transcribe → align → translate
omnicaptions transcribe video.mp4
omnicaptions LaiCut video.mp4 video_GeminiUnd.md -o video_LaiCut.srt
omnicaptions translate video_LaiCut.srt -l zh --bilingual

---

Credits: @dotey for the transcription prompt | Built on lattifai-captions

MIT License