# OpenShorts

> OpenShorts turns podcasts, webinars and streams into 9:16 clips: faster-whisper transcribes, an LLM (Gemini by default, or a local Ollama/OpenAI-compatible model) picks 3-15 moments, then MediaPipe/YOLOv8 reframes and burns in subtitles. Self-host with Docker (8GB+ RAM, CPU or NVIDIA GPU). Core is MIT; the cloud/ billing folder is source-available. A paid hosted version runs at openshorts.app. Also bundles AI UGC actor videos (paid fal.ai) and a YouTube title tool. A GitHub fork of Autocrop-vertical, but far diverged. No per-clip virality score.

- Repository: https://github.com/mutonby/openshorts
- Official docs: https://github.com/mutonby/openshorts#readme
- Checked: 2026-10-15

## What it replaces
- Opus Clip: AI picks highlight clips from a long video, Auto-reframe to vertical 9:16 (speaker or face tracking), Animated burned-in captions on clips, Schedule and publish clips to social platforms (partial). Covers the core clip pipeline and scheduled posting (via Upload-Post); no per-clip virality score; feature parity not checked.

## Facts, with sources

| Question | Status | Answer | Source | Checked |
|---|---|---|---|---|
| Runs on my machine? | Confirmed | Yes, self-hosted with Docker Compose; a paid hosted version also exists | https://github.com/mutonby/openshorts#readme | 2026-10-15 |
| Code licence | Confirmed | MIT for the core app; the cloud/ folder (billing, hosted infrastructure) is source-available under a commercial licence | https://github.com/mutonby/openshorts#license | 2026-10-15 |
| Model weights licence | Confirmed | No separate model weights; downloads third-party models (faster-whisper, YOLOv8, MediaPipe) and calls the LLM you choose | https://github.com/mutonby/openshorts#tech-stack | 2026-10-15 |
| Commercial use | Confirmed | Yes for the MIT core; the cloud/ billing folder may not be resold as a hosted service. Licences of bundled models (e.g. ultralytics YOLOv8) not checked | https://github.com/mutonby/openshorts#license | 2026-10-15 |
| Paid services needed? | Partial | Clipping can run with a local LLM and no paid key; Gemini is the default and is needed for auto-layout and silent videos; posting, dubbing and AI actors use paid APIs | https://github.com/mutonby/openshorts#6-run-without-a-google-key-local-llm-optional | 2026-10-15 |
| Hardware | Confirmed | 8GB+ RAM; CPU works (5-8 min per 8-min video); optional NVIDIA GPU, 8GB VRAM enough for large-v3-turbo | https://github.com/mutonby/openshorts#readme | 2026-10-15 |
| Operating systems | Partial | Runs in Docker; Linux and Windows (Docker Desktop with WSL2) are documented; macOS not mentioned | https://github.com/mutonby/openshorts#5-gpu-acceleration-optional-nvidia | 2026-10-15 |
| Setup effort | Confirmed | Medium, docker compose up --build, then enter API keys in the dashboard | https://github.com/mutonby/openshorts#getting-started | 2026-10-15 |

GitHub: 6693 stars, 1496 forks, last push 2026-10-16, archived: no.
---
Source: RepoQuarry (https://repoquarry.dev/projects/openshorts). Data licence: CC BY 4.0. Quotes are from project documentation and belong to their authors.