Built because the bottleneck was never the idea.
Teams do not run out of things to say. They run out of footage. Every brand we watched had a backlog of angles worth testing and a production process that could deliver one of them a month.
Generating a clip is solved. Producing a piece is not.
Plenty of tools can make eight seconds of video. Very few can make the second shot look like it belongs with the first. The interesting problems are all downstream of generation: identity that survives a cut, a product that stays the same object, a plan that knows what each shot is for.
Three habits that shape the product.
Verify against reality
A prompt change is not done because it compiled. It is done when a real generation was run and the frames were actually looked at. Most of what this product knows about directing was learned that way.
Name the limits
Every capability page here ends by saying where it falls short. Full body identity drifts more than headshots. One composed plate serves one aspect ratio. Saying so is cheaper than a customer discovering it.
Never destroy
Edits create versions, removals archive, and superseded images stay in history. You can always get back to what you had.
Real output, published and countable.
The showcase is not a highlight reel assembled by hand. It is 8 directed sequences and 6 clips generated by the product, published through a script, with every number on this site derived from that same manifest so the copy cannot drift ahead of what actually exists.
Judge it on the output.
The free tier exists so you can, before deciding anything.