# Rescript

> An open-source, transcript-based video and audio editor that runs fully offline in your browser — no uploads, no accounts, no server.

Rescript is a browser-based, transcript-driven media editor built by [@wassgha](https://x.com/wassgha). Drop in a video or audio file and it is transcribed locally using Whisper, producing per-word timestamps and speaker labels. Delete words in the transcript and the corresponding clip is cut from the media — then export the final cut as MP4 or M4A, all without your file ever leaving your device.

## What It Is

Rescript is an open-source tool that lets you edit video and audio the way you edit text. It sits entirely in the browser, using WebAssembly and WebGPU to run AI transcription and media processing on-device. The core workflow is: transcribe → edit by deleting words → export a frame-accurate cut. It targets podcasters, interviewers, and anyone who wants to trim filler words or restructure spoken content without a traditional timeline editor.

## How the Editing Workflow Works

1. **Extract** — ffmpeg.wasm decodes the audio track to mono 16 kHz PCM in-browser.
2. **Transcribe** — Whisper runs in a Web Worker with word-level timestamps; pyannote assigns a speaker label to every word. Users can choose Whisper Base (~200 MB) or Whisper Small (~600 MB).
3. **Edit** — Deleting words creates "cut ranges" that the preview player skips in real time; a waveform timeline shows cut regions in red. A one-click "Remove fillers" button cuts every detected "um", "uh", and similar filler in one pass.
4. **Export** — Kept ranges are trimmed and concatenated via an ffmpeg filter graph and re-encoded to `libx264`/`aac`, producing word-accurate MP4 or M4A output.

## Privacy-First Architecture

Rescript's defining constraint is that no media ever leaves the device. There is no server, no authentication, and no upload step. AI models (Whisper and pyannote) are downloaded from Hugging Face Hub on first use and then cached in browser storage, so after the initial model fetch the entire tool — transcription, editing, and export — works fully offline. The only third-party request the app makes is anonymous Google Analytics, which fails silently when offline.

## Technology Stack

| Layer | Technology |
|---|---|
| App framework | Next.js + React + TypeScript + Tailwind |
| Transcription | transformers.js running `whisper-base_timestamped` or `whisper-small_timestamped` (WebGPU with WASM fallback) |
| Speaker diarization | `pyannote-segmentation-3.0` (ONNX) |
| Media processing | ffmpeg.wasm (multi-threaded) |
| State management | zustand |

## Browser Support and Requirements

A Chromium-based browser is recommended. The app requires `SharedArrayBuffer`, which is served with COOP/COEP headers. WebGPU is used for inference when available, with a WASM fallback for browsers that do not support it.

## Current Status

The repository was created in late July 2026 and had accumulated 177 stars and 15 forks within days of launch, with active pushes as recently as July 27, 2026. The project is MIT-licensed and actively maintained by its author on GitHub.

## Features
- Offline, in-browser transcription using Whisper Base or Whisper Small
- Per-word timestamps and speaker diarization (pyannote)
- Word-level editing — delete words to cut corresponding media clips
- One-click filler word removal (um, uh, etc.)
- Waveform timeline with word labels, cut regions, playhead, and zoom
- Live preview that skips cuts in real time
- Frame-accurate MP4 (video) and M4A (audio) export via ffmpeg.wasm
- Supports video (MP4, WebM, MOV) and audio (MP3, WAV, M4A) input
- No uploads, no accounts — all processing on-device
- Models cached in browser storage for fully offline use after first run

## Integrations
Hugging Face Hub (model download), ffmpeg.wasm, transformers.js, WebGPU, WebAssembly (WASM)

## Platforms
WEB

## Pricing
Open Source

## Version
main

## Links
- Website: https://wassgha.github.io/rescript/
- Documentation: https://github.com/wassgha/rescript
- Repository: https://github.com/wassgha/rescript
- EveryDev.ai: https://www.everydev.ai/tools/rescript-video-editor