Tencent has released Hy-Nomas, an open model that generates a video world you steer while it runs.
Generation happens in segments, one after another. Each segment takes either a text prompt or a camera path. Tencent says the model keeps a lifelong KV cache so the scene stays consistent across turns.
The inference path is distilled. It uses 8 denoising steps and a finetuned PixelUMM decoder for 32-channel DCAE-V latents. The Diffusers example makes two five-second segments, 240 frames at 24 fps, at 1280 by 704.
The README says Hy-Nomas runs inference on a single NVIDIA GeForce RTX 4090. That config stores the DiT weights and the KV cache in INT8. It moves the text encoder, DiT and VAE to CPU between stages. Sequence parallelism is off.
Here is my read. The headline feature and the single-card config pull against each other. The feature list sells a lifelong KV cache that holds the scene together across turns. The 4090 config bounds that same cache with a sliding window. The README's words are "instead of letting the KV cache grow with every segment". So the cheapest way to run this is also the way that forgets soonest. How far that window shrinks before the world drifts is the thing I would test first.
github.com/Tencent-Hunyua…
Generation happens in segments, one after another. Each segment takes either a text prompt or a camera path. Tencent says the model keeps a lifelong KV cache so the scene stays consistent across turns.
The inference path is distilled. It uses 8 denoising steps and a finetuned PixelUMM decoder for 32-channel DCAE-V latents. The Diffusers example makes two five-second segments, 240 frames at 24 fps, at 1280 by 704.
The README says Hy-Nomas runs inference on a single NVIDIA GeForce RTX 4090. That config stores the DiT weights and the KV cache in INT8. It moves the text encoder, DiT and VAE to CPU between stages. Sequence parallelism is off.
Here is my read. The headline feature and the single-card config pull against each other. The feature list sells a lifelong KV cache that holds the scene together across turns. The 4090 config bounds that same cache with a sliding window. The README's words are "instead of letting the KV cache grow with every segment". So the cheapest way to run this is also the way that forgets soonest. How far that window shrinks before the world drifts is the thing I would test first.
github.com/Tencent-Hunyua…
Oct 24, 2026 · 8:30 PM UTC
1
0
1
145
@AIQuanting lifelong kv cache for a coherent video world? that’s wild, no more wonky scene jumps between clips.
0
0
0
2