Skip to contentRunning now · Media · automated pipeline
Loopframe
A video pipeline: write a few sentences and it storyboards them, renders every scene, records the voice-over, times the edit, adds captions and hands back a vertical MP4. No editing by hand.
Try it here
Open a sample video and inspect the stages that produced it.
The problem
Short-form video made by hand takes hours per clip: writing, finding images, recording a voice-over, timing the cuts and captioning them.
What it does
Each run goes through a queue to a worker that takes the script through eight stages: storyboard, images, voice-over, timing, edit, captions, render and a publishing pack with a thumbnail, title and description. Every stage is logged, and a failed run resumes from the stage that failed instead of starting again.
The public version renders its scenes with a built-in renderer and makes no paid AI calls; a client build calls an image API at that step.
What is in it.
- 01Script to storyboard
- 02Scene images in one look
- 03Neural voice-over, timed to the scenes
- 04Automatic edit with motion and crossfades
- 05Captions timed to the voice
- 06Every stage logged and resumable
Built with
- Python
- FastAPI
- FFmpeg
- Redis
- React