Models · 12 October 2026 · 1 min read
Helix World: a video scene you can walk through, with sound
Previewed on 12 October 2026: one picture becomes a scene you move through with the keyboard, and the sound turns with the camera.
By The Crewnema team
Noiz AI and the Hong Kong University of Science and Technology released a preview of Helix World on 12 October 2026. It is a world model: rather than rendering a fixed clip, it keeps generating video as you move through the scene. These are the points from their announcement.
What Noiz AI announced
- Video and spatial stereo sound generated together from one starting image, with an optional text prompt.
- Steering in real time: the WASD keys move the camera and the arrow keys look around, and the sound shifts as the camera turns.
- Output at 768 by 512 and 24 frames a second, generated in chunks of about 1.3 seconds, with stable 30-second runs reported.
- Real-time speed measured on one data-centre GPU (an NVIDIA H800).
- A browser demo, code on GitHub and a preview checkpoint on Hugging Face.
Our view
This is research, and it is interesting for one reason: sound that follows the camera. Most video models add sound that ignores where you are looking. For adverts it is not ready, and is not trying to be: the resolution is low, and an advert is a planned sequence of shots rather than a place you wander through. Where it could matter later is a virtual walk-through of a shop, a flat or a restaurant made from a few photos.
Side by side
In Crewnema
Helix World is not on Crewnema’s roster. The nearest thing a business can do today is turn real photos of its place into the sets of a video: how photos become locations.
Sources
Make your first video
Crewnema is in private beta and lets businesses in a few at a time. Request access and we will write when yours is open.