Prompts are dead: now you just act out your AI video on camera

Startup Simulon has shipped Reality-Aware Rendering on iPhone, and it flips the usual AI video workflow on its head. No text prompts. You use your phone camera to drop a virtual character or object into your room or backyard, hit record and just act out the scene: move, gesture, talk.

Once you stop recording, the model reads the whole take: what you did, what you said and what is around you. The characters respond to all of it, and the final clip is regenerated from the recording. Simulon co-founder Divesh Naidoo says the result is a new performance of the scene, not a copy of the footage. Your raw take works as both an audition and the instructions.

The team released a demo short and says it was made in an afternoon with zero prompting: just real footage, 3D assets and the new renderer. The end of the clip shows what was actually in front of the camera.

There are caveats. It is not real time yet, the result is rendered after the shot. Asked whether it will work live, Naidoo replied "this is where it's going". The feature is available now on iPhone, with Android coming later.

Prompt engineers may want to sit down. Now you don't describe the scene, you perform it.