Tencent Open-Sources Hy-Nomas, an Interactive Video World Model with Multi-Turn Control
Tencent-Hunyuan · hf · 2026-10-28
Tencent Hunyuan's visual generation team released Hy-Nomas, a video world model (8B diffusion backbone) for free-form interactive world generation.
- Capabilities: follows free-form text instructions and continuous camera control across multi-turn interactions, generating variable-length segments conditioned on current inputs and cached history.
- Key design: structured video captions + action-aware causal attention — descriptive context stays accessible across segments while action instructions are restricted to their intended windows; a vision-language Prompt Enhancer translates user requests and multimodal action histories into structured conditions.
- Training: multi-stage curriculum (causal generation, text interaction, camera control, model-generated histories) followed by diffusion distillation.
- Open release: weights, code, benchmarks and a lightweight adaptation workflow are public; human evaluation favors it over existing world models.
More from Multimodal
- Hands-on with Baidu's Unlimited OCR 2: whole docs in one pass, no page limit — Baidu_Inc · 2026-10-28
- Agentic video editing goes pro: senior editors 5x output, costs to collapse — dotey · 2026-10-28
- Alibaba's OvisOCR3: a 0.8B model tops OmniDocBench at 96.74 via multi-teacher distillation + RL — ATH-MaaS · 2026-10-28
- CodeCap-Bench: executable code beats natural language as video captions — Jiafu Tang · 2026-10-28
- Moving Atoms Engine turns real-world spaces and cities into high-resolution 4D models for sim — Scobleizer · 2026-10-28
- QuiverAI Launches Arrow 2 Vector Model on FLORA with Layered SVG Editing — andpoul · 2026-10-28