Hy-Nomas is an open video world model from Tencent: it builds a scene one short video piece at a time, and you can steer each new piece with text or camera moves.
In plain words
Here a world model means a model that makes a video world you can keep changing while it plays. Hy-Nomas creates video in segments, and before each new segment you can add objects, switch the style or scene, describe actions or move the camera. A memory of earlier segments, called a KV cache, keeps the world consistent. A cut-down setup runs on one NVIDIA GeForce RTX 4090 graphics card. Tencent also released Hy-Nomas-Bench, 280 test cases for judging such models. The code uses the Apache 2.0 licence.
A quick example
Imagine you type "a quiet beach at sunset". Hy-Nomas makes the first few seconds, then you ask for a sailboat and swing the camera left, and the next segment adds the boat while the beach stays the same.
Where you'll meet it
- Making interactive video worlds steered by text and camera paths.
- Robot simulation, with an add-on trained on robotic-arm data.
- Testing interactive world models with Hy-Nomas-Bench.
Related
Sources
- Docstencent/Hy-Nomas, Tencent (Hugging Face) · read 24 Oct 2026
- DocsTencent-Hunyuan/Hy-Nomas, Tencent (GitHub) · read 24 Oct 2026
Checked by WhatIsAI against the sources above · 24 Oct 2026. A full five-minute explainer with a diagram may follow. Spotted an error? Report it.