Features
Real-time video generation at 25 fps
Keyboard and mouse input control
Auto-regressive diffusion-based image-to-world framework
Long video generation with temporal coherence
Open-source code and models (MIT license)
Three pretrained scenes: universal, GTA driving, TempleRun
Demo utilities and configuration files
Hugging Face model integration
Local inference on Linux with NVIDIA GPU
Image-to-image editing and inpainting
Stable Video: short video clips from text or image
Stable Audio 3.0: open-weight music & sound effect generation (May 2026)
3D model generation from text or images
Brand Studio: managed enterprise creative production platform
Curated Model Routing across multiple providers
Precision Inpainting and Product Insertion tools
Brand Central for custom models and campaign guidelines
Self-hosted deployment for data privacy
Cloud API for integration without infrastructure
Batch processing for high-volume assets
Open-weight models for community experimentation
Sketch-to-image and recoloring tools
Producer Mode for planning multi-step workflows
