jianshuo/claude-skills — explained in plain English
Analysis updated 2026-05-18
Turn raw video into subtitled, translated, and dubbed footage using a chained set of Claude Code skills.
Automatically align and cut between multiple camera angles shot at the same event based on who is speaking.
Split a long video into shorter topic based clips and reframe them between landscape and portrait.
Draft, illustrate, and publish an article to WeChat with an automatically generated cover image.
| jianshuo/claude-skills | jeremyiv/elastic-streets | opsintech/opsintech-platform | |
|---|---|---|---|
| Stars | 34 | 34 | 34 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 3/5 | 3/5 | 4/5 |
| Audience | vibe coder | data | ops devops |
Figures from each repo's GitHub metadata at analysis time.
Requires Claude Code installed first, then copying or installing individual skill folders.
claude-skills is a collection of 13 skill packages the author uses daily inside Claude Code, an AI coding assistant from Anthropic. A skill in this context is a folder with a SKILL.md file and supporting scripts that Claude automatically loads when a user's request matches certain trigger words, so the user does not need to explain the whole workflow every time. Most of the skills here center on a video production pipeline: turning raw footage from multiple cameras into a finished, translated, and dubbed video ready to upload. One group covers transcribing audio into subtitles, translating those subtitles into another language with proper punctuation based line breaks, generating text to speech dubbing in the target language, and finally burning the subtitles into the video while mixing in the dubbed audio, all in a single encoding pass to avoid losing video quality. A separate orchestrator skill can run all four of these steps together for a full localization job, or each one can be called on its own for a single step. Another group of skills handles multi camera footage. One aligns audio from several cameras shot at the same event by comparing their sound waves to work out the time offset between them, and stores that offset in a small sidecar file next to each original clip rather than re-encoding the footage. Another skill then uses those alignment files to automatically cut between camera angles based on which one is talking loudest at any moment, producing a single edited video. Additional skills split a long video into shorter topic based clips, add overlays like covers and captions, convert between landscape and portrait framing while tracking the speaker, and upload finished videos to YouTube in bulk. There is also a skill for writing and publishing articles to WeChat, a Chinese messaging and content platform, which lightly edits a draft's wording, generates a cover image and explanatory illustrations, and uploads the draft through a separate publishing tool. A few remaining skills cover project auditing and personal reflection routines. The skills follow the SKILL.md format that Anthropic published and that OpenAI's Codex CLI later adopted, so besides Claude Code they are described as also working with Codex CLI, with support for Cursor and Gemini CLI in progress. Installing a skill is a matter of copying its folder into Claude Code's skills directory, or installing it through a third party skill marketplace.
A set of 13 Claude Code skills for a video production pipeline (transcribe, translate, dub, multicam sync and edit, reframe, upload) plus a WeChat article publishing skill.
Mainly Python. The stack also includes Python, Claude Code, FFmpeg.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly vibe coder.
This repo across BitVibe Labs
Verify against the repo before relying on details.