Hey PH! I'm the solo developer behind ClipForge. Here's the backstory: I edit long-form content (podcasts, streams) and the clipping workflow was always: upload 5GB to a cloud tool → wait 20 minutes → get watermarked clips → pay $30/mo to remove watermarks. It felt wrong that MY footage had to leave MY computer just to get cut into clips. So I built ClipForge to run entirely on your own machine. The Whisper transcription model downloads once (~500MB) and then everything works offline. Your raw video never touches a server. What's under the hood: - faster-whisper (CTranslate2) for transcription — C++ inference, ~2x realtime on a decent CPU - 6-feature heuristic scorer that ranks every moment without needing an API key - Optional LLM scoring via BYO key (OpenAI/Anthropic/DeepSeek) — you pay your provider, I never mark up tokens - ONNX face detection for face-aware 9:16 cropping - ffmpeg pipeline for export with word-level timed captions Current state: v0.4.0, 780+ tests passing, on my 5th iteration. Built with Python, PySide6, and a lot of late nights. What I'd love feedback on: - Would you use this instead of Opus Clip / 2short? What would make you switch? - The trial is 60 minutes / 15 clips — is that enough to evaluate? - Mac version: how many of you would use this on macOS? Ask me anything about the build, the pricing model, or the face tracker — happy to get technical.

ClipForge - Turn long videos into vertical clips

Replies