marcelpanse
Featured
youtube-guitar-tab-parser
A CLI tool that downloads YouTube guitar lesson videos, uses Claude vision to detect and extract the tab region from sampled frames, de-duplicates lines by bar number, and assembles them into a PDF.
Why it is worth attention
It combines state-of-the-art on-device transcription (Whisper) with speaker diarization and word-level editing, all without sending your media to any server, making it a rare privacy-first alternative to cloud-based editing tools.
Who it is for
- Video editors and content creators who want a private workflow
- Podcasters and audio editors looking to quickly remove fillers and mistakes
- Journalists and interviewers needing fast transcript-based cutting
- Privacy-conscious users who cannot upload media to online services
Use cases
- Editing a recorded interview by deleting filler words or off-topic sections in the transcript
- Removing 'um', 'uh', and other filler sounds from a podcast or voice note with one click
- Importing an existing SRT/VTT transcript to skip on-device transcription and edit directly
- Exporting a frame-accurate MP4 or M4A of the final cut without leaving the browser
Strengths
- Fully private — no uploads, no server, no authentication required; media stays on your device
- Word-accurate editing with live preview and timeline showing cut regions in real time
- Supports both video and audio files, with optional speaker diarization
- Open-source (MIT) and built with modern web standards (WebGPU, SharedArrayBuffer, ffmpeg.wasm)
Considerations
- First-time use requires downloading Whisper models (~200-600 MB) and a speaker model, which may take time
- Requires a Chromium-based browser with SharedArrayBuffer support; not compatible with all browsers
- Editing is limited to transcript-based cuts — you cannot trim or edit media without generating or importing a transcript
README quick start
Rescript
Edit video and audio like you edit text — fully offline, in your browser.
✨ Try it now: wassgha.github.io/rescript
Rescript is an open-source, transcript-based media editor. Drop in a video or audio file and it is transcribed locally with per-word timestamps and speaker labels. Delete words in the transcript and the corresponding clip is cut from the media. Export the final cut — without your file ever leaving your device.
- 🔒 Private by design — no server, no auth, no uploads; all media processing happens on-device
- 📝 Word-level editing — select words, press ⌫, the cut follows the text
- 📥 Import your own transcript — skip Whisper and edit with an SRT, VTT, or JSON caption file
- 🧹 Filler removal — one-click cut of "um", "uh", and similar fillers
- 🗣️ Speaker diarization — the transcript is grouped by speaker
- 🎬 Timeline — waveform, word labels, cut regions, playhead, zoom
- ⚡ Live preview — playback skips your cuts in real time
- 📦 In-browser export — frame-accurate MP4 (video) or M4A (audio) with ffmpeg.wasm
- 🎧 Audio files — edit podcasts, voice notes, and interviews the same way as video
Stack
| Piece | Tech |
|---|---|
| App | Next.js + React + TypeScript + Tailwind |
| Transcription | transformers.js running whisper-base_timestamped or whisper-small_timestamped (WebGPU with WASM fallback) in a Web Worker |
| Speaker labels | pyannote-segmentation-3.0 (ONNX) |
| Media processing | ffmpeg.wasm (multi-threaded) for audio extraction and export |
| State | zustand |
Development
npm install # also copies ffmpeg/onnxruntime WASM into public/vendor
npm run dev # dev server
npm run build # production build
npm run lint # eslint
Open http://localhost:3000 and drop in a video with an audio track.
Note on "offline": the AI models (Whisper Base ~200 MB, or Small ~600 MB, plus a small speaker model) are downloade
Description
🎬 Open source, transcript-based video/audio editor that lives in the browser.
Related repositories
Similar projects matched by category, topics, and programming language.
pireel
pireel
Pireel Studio is an open-source, backend-free AI video editor for talking-head videos that runs entirely in the browser with no account or server required.
pengchujin
jzsub
JZSub is a Codex skill that downloads videos from popular platforms, generates bilingual subtitles via GPT, and burns them into MP4 files with high quality and customizable delivery options.