If you're posting video for your business, you already know short-form clips are where the attention is. The problem: turning a 20-minute recording into a punchy 60-second reel takes real time, and the tools that do it automatically (OpusClip, Munch, Descript's clip feature) all run $30 to $100 per month with watermarks until you pay up.
SupoClip launched on Product Hunt today, and it does the same job for nothing. It's a fully open-source tool by indie developer FujiwaraChoki that takes a long video, finds the best moments using AI virality scoring, adds captions automatically, handles face tracking, and exports vertical 9:16 clips ready for TikTok, Instagram Reels, or YouTube Shorts. The project is on GitHub, AGPL-3.0 licensed, and self-hosted via Docker Compose.
The part that makes it genuinely free: it works with a local Ollama model for the AI scoring, which means zero per-clip cost once it's running. For a shop posting weekly recaps, tutorial recordings, or event coverage, this could replace a paid subscription entirely.
Try this this week
- Clone and start it:
git clone https://github.com/FujiwaraChoki/supoclip && cd supoclip && docker compose up - Install Ollama (ollama.com) and pull a lightweight model like
llama3.2orgemma2:2bfor virality scoring. Both run fine on a basic laptop or a cheap VPS. - Drop in a real recording. A recent client event video, a staff intro, or a recorded webinar is the perfect first test. Set your target clip length (30s, 60s, or 90s) and let it find the moments.
- Review the exports. Face tracking keeps the subject centered in the 9:16 frame. Check a few clips, adjust the minimum virality threshold if you want more or fewer results.
- Post one this week. Don't wait for perfect. Clip, caption, post. See what gets traction before you tune the settings.
The bigger picture
The paid clip-tool market is built on one assumption: editing for social media is too fiddly for small teams to own themselves, so they will pay monthly forever. SupoClip punches a hole in that. You run it yourself, you own the data, and the per-clip cost is zero. That's the pattern build9 keeps seeing across AI tooling: the open-source version catches up faster than the vendors expect, and small teams that run their own stack stop paying subscription rent. Video clipping just joined that list.