DimCut
May 18, 2026 · View on GitHub
English ｜ 中文
DimCut is a novel editing interaction design that folds the 1D timeline into multiple rows, integrating text, audio, and visuals — multidimensional information at a glance.
I believe high-density information improves the editing experience and efficiency, especially for knowledge-based videos such as talks, interviews, podcasts, and more.
DimCut is built on web technologies. All processing happens locally in the browser — your files are never uploaded to any server.
Features
- 2D Multi-row Timeline — The 1D timeline is folded into rows, giving you a full overview on one screen. No more endless panning and zooming.
- ASR Text Track — Speech transcripts are aligned alongside media. Copy, delete, or reorder text — that equals editing the video.
- Text-first Editing — When content is driven by speech (lectures, interviews, podcasts), text-first editing is simply the fastest way.
- Video Thumbnails — Extracted via mediabunny CanvasSink, lazily loaded with IntersectionObserver.
- Audio Waveforms — Decoded once and cached as raw peaks, rendered as HiDPI bar-style waveform on Canvas.
- Preview Playback — Built-in player with playhead sync and automatic skip over deleted ranges.
- Video Export — Trim or export clips with streaming writes to avoid OOM. All done in-browser.
- Offline Web App — No server, no upload, no account. Everything runs in your browser.
Quick Start
# Clone the repo
git clone https://github.com/hughfenghen/dimcut.git
cd dimcut
# Install dependencies
pnpm install
# Start dev server
pnpm dev
Open http://localhost:5173 to see the landing page and interactive demo.