2.8B-parameter open-source interactive world model that turns an image, prompt and WASD input into video on one GPU. A distilled variant makes a minute-long clip in 34 seconds on an RTX 5090; weights are coming soon.
NVIDIA's 2.8B-parameter world model for minute-long, camera-controlled 720p video generation on a single GPU.
Updated EU status checked Last shown
SANA WM at a glance
- What is SANA WM?
- SANA WM is a model from NVIDIA. NVIDIA's 2.8B-parameter world model for minute-long, camera-controlled 720p video generation on a single GPU.
- Who makes SANA WM?
- SANA WM is made by NVIDIA.
- Does SANA WM work in the EU?
- Unverified. No explicit evidence found. Weights were reported as not yet released, and no regional restrictions were seen on the project page. Last verified 14 Nov 2026.
- How much does SANA WM cost?
- Pricing unknown. Reported as open-source with weights coming soon; no pricing information found.
- Where was SANA WM shown?
- In 1 creator video by AI Search. Most recently by AI Search on 17 May 2026, at 6:38 in “Real gundams, top 3D generator, open-source world models, ChatGPT updates, new TTS: AI NEWS”.
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
SANA-WM is an efficient world model from NVIDIA Labs (the SANA project) for minute-scale, camera-controlled video generation at 720p. According to a creator review, it has 2.8B parameters and turns an input image, a text prompt and WASD-style camera/movement input into an interactive video sequence.
It is designed to run on a single GPU. A distilled variant is reported to generate a minute-long clip in about 34 seconds on an RTX 5090. The project page describes it as an efficient minute-scale world model.
Weights were reported as "coming soon" at the time of the source material, so availability of downloads, license terms and code repository are unconfirmed. It is aimed at researchers and developers working on world models, interactive simulation and game-like video generation.
- 2.8B-parameter world model
- Minute-scale 720p video generation
- Camera control via WASD input plus image and text prompt
- Runs on a single GPU
- Distilled variant: ~34 s for a one-minute clip on RTX 5090 (reported)
As seen on YouTube
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Similar AI tools
Closest neighbours by meaning, not just by category.
Sana Video 2
Model · Nvidia
NVIDIA's 5B and 14B hybrid linear-attention video diffusion transformers generating up to 720p on a single GPU.
Hy-Nomas
Model · Tencent
Tencent Hunyuan interactive world model generating real-time, controllable 720p video with memory for consistency.
Cosmos 3
Model · NVIDIA
NVIDIA's open omni-model family for physical AI: world simulation, reasoning and action generation for robots and AVs.
WorldCrafter
Model · Tencent ARC Lab
Video world model from Tencent ARC Lab that turns an image or text into camera-controlled explorable scenes with 3D-aware memory.
GWM Worlds 2
Model · Runway
Runway's real-time interactive world model generating 720p/24fps video with audio, steered by text actions and camera input.
SolarWM
Model
Open framework and dataset for turning video diffusion models into real-time, long-horizon interactive world models.