NVlabs Launches SANA: Open-Source AI Framework Generates 4K Images and Video Up to 100x Faster Than Competitors
Summary
NVlabs launches SANA, a groundbreaking open-source AI framework capable of generating stunning 4K images and video up to 100 times faster than competitors like Flux-12B, with multiple model variants, flexible deployment options including low-VRAM laptop support, and a fully open-sourced codebase under the Apache 2.0 license.
Key Points
- NVlabs releases SANA, an open-source efficiency-focused framework for high-resolution image and video generation, featuring models up to 4K resolution that run up to 100 times faster than competing systems like Flux-12B.
- The framework includes multiple model variants — SANA, SANA-1.5, SANA-Sprint, SANA-Video, and the newly released SANA-WM — leveraging key innovations such as Linear Attention, 32x image compression via DC-AE, and Block Causal Linear Attention for long video generation.
- SANA supports flexible deployment options including 4-bit quantization for laptops with under 8GB VRAM, integration with Hugging Face Diffusers, ComfyUI, SGLang, and RL post-training via Cosmos-RL, with the codebase fully open-sourced under the Apache 2.0 license.