BetterClipper

Desktop video clipping tool. Paste a YouTube URL, search the transcript, scrub the timeline, and export a frame-perfect clip — at your chosen resolution.

Built with Tauri v2 (Rust) + React (TypeScript).

It works by:

1. Fetching automatic or creator-provided subtitles via yt-dlp (no video download yet).
2. Merging YouTube's word-level caption fragments into clean, searchable segments.
3. Searching the transcript in real-time or clicking segments to download a ±10s preview.
4. Scrubbing the preview with In/Out markers on a zoomable timeline.
5. Exporting a frame-perfect clip re-encoded via ffmpeg at your chosen resolution.
6. Falling back to local whisper.cpp transcription when no subtitles are available.

Quick start

git clone https://github.com/SecareLupus/YoutubeClipper.git
cd YoutubeClipper
pnpm install
pnpm tauri dev

pnpm tauri dev compiles the Rust backend, starts the Vite dev server, and opens the desktop window — all in one command.

Requirements

Runtime sidecars (must be on your PATH in dev mode):

Binary	Purpose	Install
yt-dlp	Video download + subtitle extraction	pip install yt-dlp or github.com/yt-dlp/yt-dlp
ffmpeg	Audio demuxing + frame-perfect re-encoding	sudo apt install ffmpeg (Linux), brew install ffmpeg (macOS), or ffmpeg.org
whisper-cli	Local transcription fallback	Build from ggerganov/whisper.cpp

Build dependencies (for compiling from source):

• Rust (stable toolchain)
• pnpm (Node.js package manager)
• Linux: libwebkit2gtk-4.1-dev, libgtk-3-dev, libappindicator3-dev, and related Tauri system libraries

See the Tauri v2 prerequisites for platform-specific setup.

Usage

1. Launch with pnpm tauri dev
2. Paste a YouTube URL (or any site yt-dlp supports) and click Load
3. The transcript loads in the right panel — search it or scroll through segments
4. Click a segment (or search match) to download a ±10 second preview
5. The video player loads with In/Out markers set to your selection
6. Drag the markers or use keyboard shortcuts to fine-tune the cut
7. Select a quality preset and click Export Clip to save the final .mp4

Keyboard shortcuts

Key	Action
Space	Play / pause
, / .	Frame back / forward (1/30s)
← / →	Skip ±5 seconds
Ctrl+← / →	Jump to nearest keypoint (In, Out, start, end)
I	Set In marker at playhead
O	Set Out marker at playhead

Timeline

• Zoom bar — shows the selected clip region with draggable In/Out markers and a moving playhead
• Zoom handles — drag the left/right edges of the bar to expand or narrow the view
• Overview bar — shows the zoom window position relative to the full clip duration
• Time readout — In, Out, and Clip duration displayed above the bar

Speech-to-text fallback

When yt-dlp cannot find subtitles for a video, the app automatically falls back to local transcription via whisper.cpp:

1. Downloads the 720p video
2. Extracts a 16kHz mono WAV with ffmpeg
3. Transcribes with whisper-cli using the ggml-tiny.bin model

The model (~78 MB) is auto-downloaded from HuggingFace on first use and cached in ~/.cache/better-clipper/models/.

Architecture

┌──────────────────────────────────────┐
│              React UI                │  TypeScript + Vite
│  Video · Timeline · Search · Export  │
└──────────────┬───────────────────────┘
               │ Tauri IPC (invoke / events)
┌──────────────▼───────────────────────┐
│           Rust Backend               │  Tauri v2 commands
│  Downloader · Transcriber · VTT ·    │
│  Export · HTTP server (video serve)  │
└──────────────┬───────────────────────┘
               │ tokio::process + std::net
┌──────────────▼───────────────────────┐
│  yt-dlp    ffmpeg    whisper-cli     │  Sidecar binaries
└──────────────────────────────────────┘

Project status

Builds and runs. Core workflow complete:

• YouTube captions → transcript fetch (no video download until needed)
• Transcript search with debounced query, match highlighting, and interpolated timestamps
• Segment merging (collapses YouTube's word-level fragments)
• Time-bounded section download via yt-dlp --download-sections
• Preview clip modal — renders the exact I/O range via ffmpeg, cached by time markers
• Preview stops on last frame (no auto-close), autoplays reliably via programmatic play
• Video player with play/pause, frame-stepping, and skip controls
• Zoomable timeline with draggable In/Out markers and playhead tracking — renders immediately on section load
• Frame-perfect export via ffmpeg (libx264, ultrafast preset, CRF 18, AAC 192k)
• Multi-resolution export (Best, 1080p, 720p, 480p)
• Save dialog with suggested filename (video title + timestamp range)
• Live progress bars (download, model fetch, transcription, export encoding)
• HTML entity decoding in transcripts (& <   etc. → readable text)
• Whisper.cpp fallback with auto-downloaded ggml-tiny.bin
• Local HTTP server with Range support for <video> element playback
• Video title detection for export filename suggestions
• Section downloads keyed by time range — re-selecting a segment downloads afresh

Releases

Pre-built packages are available on the Releases page. To create a new release, run the Release workflow from the Actions tab with a version number — it tags, builds all platforms, and publishes the artifacts.

Linux compatibility

Package	Works on
.deb	Ubuntu 24.04+, Debian 13+, Linux Mint 22+, Pop!_OS 24.04+
.rpm	Fedora 39+, openSUSE Tumbleweed
.AppImage	Any Linux (self-contained, no system deps needed)

Ubuntu 22.04, Debian 12, and RHEL 9 are not supported by the .deb/.rpm — they ship libwebkit2gtk-4.0 (Tauri v2 requires the 4.1 API). Use the .AppImage on those systems.

Windows / macOS

Windows (.msi, .exe NSIS) and macOS (.dmg) packages are built from CI but not yet tested. Sidecar binaries (yt-dlp, ffmpeg, whisper-cli) are downloaded during the CI build for each platform.

License

MIT