PrimoDev/ apps

PrimeClip

AI vertical clips from long-form video, on your own computer

  • macOS
  • Windows
  • Linux
  • Free
  • Open source · MIT
  • v0.1.1

What is PrimeClip?

PrimeClip is a desktop app that turns long-form video into vertical clips for Shorts, Reels and TikTok. It transcribes with Whisper, detects scenes, uses an LLM to pick the best moments, tracks faces for the vertical crop and renders with captions in FFmpeg, all on your machine.

It's open source under the MIT license. Release v0.1.1 ships installers for macOS (.dmg), Windows (.exe and .msi) and Linux (.deb, .rpm and .AppImage) in two variants: Full, with FFmpeg and the Whisper base model bundled, and Lite, which uses your system FFmpeg.

What it does

  1. 01A full pipeline: audio extraction, Whisper transcription, scene detection, AI clip selection, face tracking, vertical crop, captions and render
  2. 02Pick your LLM: local Ollama, Claude, OpenAI, OpenRouter or any OpenAI-compatible API
  3. 03API keys stored in the OS keychain
  4. 04Import video from disk, a YouTube or Vimeo link (via yt-dlp) or Google Drive
  5. 0530 to 90 second clips, each with a score and the reason it was picked
  6. 06Caption editor with a timeline, waveform and preview
  7. 07Download each clip as an MP4
  8. 08English and Portuguese interface

FAQ

Is PrimeClip free?
Yes. It's open source under the MIT license. If you use Claude, OpenAI or OpenRouter you pay those providers for API usage directly; with local Ollama there's no API cost.
Does my video leave my computer?
The video doesn't. Audio extraction, Whisper transcription, scene detection, face tracking and rendering all run locally. For clip selection the text transcript goes to the LLM you choose: with Ollama everything stays on your machine, with a cloud provider the transcript text is sent to it.
Which operating systems does it run on?
It's a Tauri app for macOS, Windows and Linux. In v0.1.1 the Full variant is only out for Apple Silicon Macs; Lite has builds for Apple Silicon Macs, Windows (x64) and Linux (x64).
Should I download Full or Lite?
Full is for creators who want to open it and go: it bundles FFmpeg and the Whisper base model. Lite is for developers who already have the toolchain and uses the system FFmpeg. Neither bundles Ollama, so install it separately if you want a local LLM.
Is it an OpusClip alternative?
That's the idea: the repository describes itself as an OpusClip alternative, with the difference that it runs on your computer and lets you pick the AI model.
What do I need to run it from source?
Python 3.12+, uv, Node.js 20+, Rust (for Tauri) and FFmpeg 6+ with libass. Ollama is optional. Start the backend with make dev-api and the UI with make dev-desktop.

Documentation

README from the lucianodiisouza/PrimeClip repository, refreshed on every deploy.

Local-first, AI-powered video clipping tool. Turn long-form video into vertical Shorts/Reels/TikTok clips using Claude, OpenAI, OpenRouter, or local Ollama.

Features

  • Upload long-form video → automatic clip generation
  • Pipeline: audio extract → Whisper transcription → scene detection → LLM clip selection → face tracking → vertical crop → captions → motion → FFmpeg render
  • Pluggable LLM providers (Ollama, Claude, OpenAI, OpenRouter, custom OpenAI-compatible APIs)
  • Tauri desktop app (macOS, Windows, Linux) with OS keychain for API keys
  • Full and Lite distribution variants

Prerequisites

  • Python 3.12+
  • uv
  • Node.js 20+
  • Rust (for Tauri)
  • FFmpeg 6+ with libass
  • Ollama (optional, for local LLM)

Quick Start (Development)

# Install backend deps
uv sync --all-extras

# Install frontend deps
cd apps/desktop && npm install && cd ../..

# Copy env
cp .env.example .env

# Terminal 1 — API
make dev-api

# Terminal 2 — Desktop UI
make dev-desktop

API docs: http://127.0.0.1:8765/docs (when PRIMECLIP_DEBUG=true)

Which build should I download?

Variant Best for Includes
Full Creators wanting one-click setup Bundled FFmpeg, Whisper base model
Lite Developers with existing toolchain App + Python sidecar only; uses system FFmpeg

Ollama is never bundled — install separately for local LLM.

Project Structure

backend/          FastAPI sidecar + pipeline services
apps/desktop/     Tauri 2 + React UI
scripts/          Build & dependency scripts
resources/        Vendored binaries (Full builds)
outputs/          Runtime uploads, artifacts, clips (gitignored)

API

Method Path Description
GET /api/health Health + dependency status
GET /api/providers List LLM providers
POST /api/providers/test Test provider connection
POST /api/upload Upload video
POST /api/generate-clips Start clip generation job
GET /api/jobs/{id} Job progress
GET /api/clips/{video_id} List clips
GET /api/download/{clip_id} Download MP4

Testing

make test
make lint
make check-deps

Building Releases

make bundle-backend   # PyInstaller sidecar
make build-lite       # Lite desktop build
make build-full       # Full desktop build (vendored FFmpeg)

License

MIT — see LICENSE