pmarreck

Yt Transcriber — AI skill for Claude Code

AI community

CLI app- Give it a YouTube URL and you get a transcription with possible speaker identification and optional summary or translation, all thanks to open-source AI tooling and my lack of enough free tim.

How to install Yt Transcriber

This entry records only its repository, not the path inside it, so there is no exact command to give. Open pmarreck/yt-transcriber and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Yt Transcriber does

CLI app- Give it a YouTube URL and you get a transcription with possible speaker identification and optional summary or translation, all thanks to open-source AI tooling and my lack of enough free time to watch content-sparse YouTube videos

Alternatives in AI

  • Qiaomu Markdown Proxy — Fetch any URL as clean Markdown via proxy services (r.jina.ai / defuddle.md) or built-in scripts 426 ★
  • Pii Guard — 繁體中文(台灣)個人資料去識別化工具,讓業務文件可以安全地送進 AI 處理 / Reversible Traditional-Chinese (Taiwan) PII de-identification for LLM 108 ★
  • HydraTeams — Translation proxy that makes Claude Code Agent Teams model-agnostic 62 ★

README

yt-transcriber

CLI app — Give it a YouTube URL (or a path to a video or audio file) and you get a transcription with optional summary or translation and in-progress speaker identification, all thanks to open-source AI tooling and my lack of enough free time to watch content-sparse YouTube videos

features

  • transcribe YouTube videos by URL
  • output metadata about the video
  • speaker identification (probably using an LLM in conjunction with a speaker diarization library)
  • summarization via summarize (requires OPENAI_API_KEY to be set)
  • translation via translate (requires OPENAI_API_KEY to be set)
  • can use almost any audio or video format that ffmpeg can handle as input, not just YouTube URLs
  • Test suite (run it with yt-transcriber TEST or TEST=1 yt-transcriber)
  • support for other video platforms
  • convert all this to a web service or web app

Speaker identification ("diarization"), summarization and translation will probably require an API key for Claude or OpenAI and/or one from Huggingface.

installation

NEW: If you have Nix installed or are running on NixOS, just symlink `yt-transcriber`, `summarize` and `translate` to any directory (usually `~/bin` or `XDG_BIN_HOME` which is usually `~/.local/bin`) in your `PATH` and you're good to go (the last two require OPENAI_API_KEY to be defined in your environment). The shell script will automatically procure all dependencies deterministically and locally and cache them. It pins nixpkgs to `github:NixOS/nixpkgs/1b7e751db23cedcd3bf56dc5170840b8a7e752a2` and will re-exec under `nix shell` automatically (no need to run `nix develop` yourself). You can opt into GPU builds with `ENABLE_ROCM=true` or `ENABLE_CUDA=true`.

If you do not have Nix installed, I recommend using the Determinate Nix Installer from here: https://github.com/DeterminateSystems/nix-installer

If you refuse to use Nix, you can try to install the following dependencies manually,