A still image becomes a finished short video in under a minute, with no editing software involved.
Upload a still, get a short animated video back. A rendering queue underneath, one button on top.
- Fixed scope · 6 weeks
- 2024
- Media
- AI product, Web app
The situation
The client's users had product and content photography and needed short-form video. The existing options were to learn a video editor, pay someone, or buy stock motion that did not match the product. Most of them did none of the three and posted the still image instead.
What we built
One upload field and one button, with a queue and a rendering pipeline behind it that can absorb a spike without falling over. Generation models handle the motion; the platform handles everything that makes the models usable — queueing, retries, storage, delivery and a clear state for the user while they wait.
What we deliberately left out
No editing timeline. Adding trimming, layers and keyframes would have rebuilt the thing users were avoiding. The product's whole value was that it was one button.
No model selection. Users do not want to choose a model. We picked, evaluated, and kept the choice internal so it could change without anyone relearning the product.
No accounts or billing in v1. Shipped as a working generator first, to find out whether people came back, before building a way to charge them.
The result
A still becomes a finished clip in under a minute. Small brands and creators use it to produce short-form video from photography they already own, without buying stock footage or learning an editor.
Built with
- Python
- FastAPI
- Next.js
- AWS S3
- Replicate
Have something like this?
Tell me what the problem is on a 30-minute call. If it's a fit, discovery starts with a written scope and a fixed number — not an estimate that moves.