Tencent Hunyuan has added a new model called Hy‑Nomas to Hugging Face. It generates video worlds one segment at a time, letting users add objects, change style, or move the camera through multi‑turn text instructions. A distilled version runs in eight denoising steps at 720 × 1280 resolution, 24 FPS, and a single‑GPU setup works on an RTX 4090 with INT8 weights. The model and a benchmark suite, Hy‑Nomas‑Bench, are available for download along with a web demo.
Why it matters
Creators can experiment with interactive video generation on a single high‑end GPU without needing large cloud clusters.
