Udder App — Backend API

FastAPI backend for the Udder AI video montage service. Accepts a video upload, uses OpenAI and Google Gemini to identify the best clips, and returns timestamped clip data for the client to assemble.

How it works

1. Client uploads a video file
2. API extracts audio and transcribes it
3. OpenAI/Gemini analyses the transcript and video to identify compelling clip segments
4. Returns clip list (start/end times, reason for inclusion) to the client
5. Client assembles the final montage (browser uses ffmpeg.wasm, desktop uses ffmpeg)

Stack

• FastAPI + uvicorn
• OpenAI API — clip selection analysis
• Google Gemini — secondary analysis model
• ffmpeg — server-side audio extraction
• Docker + DigitalOcean App Platform

Setup

pip install -r requirements.txt
# Requires: OPENAI_API_KEY, GEMINI_API_KEY in .env
uvicorn app.server:app --reload

Or with Docker:

docker build -t udder-app .
docker run -p 8000:8000 udder-app

Related

• udder — React web frontend
• udder-demo — PyQt6 desktop client