Interactive keyboard-controlled video model runs in real time with just 0.5B parameters
multimodalart · x · 2026-07-29
The author says they are still unsure whether to call these systems “world models,” but says the demo now runs in real time.
- The model is a keyboard-controlled video model with decent quality.
- It uses only 0.5B parameters.
- The demo can run on ZeroGPU Spaces and also locally.
More from Multimodal
- Flux 3 prompt aims for a bootleg phone recording of a metal encore — fofrAI · 2026-07-29
- NVIDIA’s PDD speeds image and video generation with 4–8-step distillation — nvidia · 2026-07-29
- VisoMaster open-source tool swaps faces in images and videos with LivePortrait tuning — tom_doerr · 2026-07-29
- A generated img 2.0 meme shows Elon Musk talking with Sam Altman — SkyNo7576 · 2026-07-29
- ComfyUI 0.29 adds streaming video, GPT 5.6, Claude Opus 5, and new partner nodes — pytraveler · 2026-07-29
- Creator says an album was made end to end with AI using Suno — Kyrannio · 2026-07-29