We (@lawrence_cjs, @yuyangzhao_ , @shanasaimoe) from the SANA team just posted a blog on the core of Linear Attention: how it achieves infinite context lengths with global awareness but constant memory usage! We explore state accumulation mechanics, the evolution from Softmax to

How Linear Attention and Softmax Attention differ in compute and KV-Cache for LLMs and long-video generation. Let's start with this blog. https://t.co/Ja5El08muf

Sora 2 is amazing!, But AI video generation inference speed is too slow. Try our Deep Compression Autoencoder + Linear Attention! 🚀🔥 https://t.co/ooNowz8HH7 https://t.co/PU8oUI2hsU

Thanks so much @_akhaliq for sharing our recent work. Our homepage is here:

Changing the autoencoder in latent diffusion models is easier than you think. 🚀 Introducing DC-Gen – a post-training acceleration framework that works with any pre-trained diffusion model, boosting efficiency by transferring it into a deeply compressed latent space with

We release DC-VideoGen, a new post-training framework for accelerating video diffusion models. Key features: 🎬 Supports video generation up to 2160×3840 (4K) resolution on a single H100 GPU ⚡ Delivers 14.8× faster inference than the base model while achieving comparable or

🚀 SANA-Video: Linear Attention + Constant-Memory KV Cache = Fast Long Videos 💥 Key Features 🌟 🧠 Linear DiT everywhere → O(N) complexity on video-scale tokens 🧰 Constant-memory Block KV cache → store cumulative states only (no growing KV) 🔄 🎯 Temporal Mix-FFN + 3D RoPE

Finally, 36s for 5s 720p on H100; 4× speedup vs vanilla attention at 720p 29s on RTX 5090 with NVFP4 (2.4x faster) Fixed VRAM vs sequence length; strong text–video alignment

3. Temporal Mix-FFN+3D RoPE → local fidelity + temporal coherence 🎯 4. AR block training with Self rollout → minute-length generation 📊

2. Constant-Memory Block KV cache→cumulative states only (no growing KV) 🔄

Keys🌟 1. Linear DiT everywhere → O(N) complexity on video-scale tokens

🚀 SANA-Video: Linear Attention + Constant-Memory KV Cache = Fast Long Videos 💥 It's time for a new SANA family member! Links 🌐 📖 Paper: https://t.co/snV4bF8jUM 💻 Project Page: https://t.co/9WZIp7ryX6

Explore recent work from our team. Long-Live generates minute-length videos and interacts as you want with real-time fast speed! Very cool project. 🎉

Explore Deep Compression Autoencoder (DC-AE) 1.5 with higher token compression ratio (64x) for faster visual generation:

The best few-step sampling model across the speed-memory frontier? 😱 Introducing SANA-Sprint in collaboration with the great SANA team! Beyond the results, perhaps more importantly, the work is about the recipe of SANA-Sprint. Code & model will be open ❤️ Let's go ⬇️

SANA-Sprint One-Step Diffusion with Continuous-Time Consistency Distillation

Explore our one-step diffusion model, SANA-Sprint. Very fast:

Still think consistency models are bad at scale? In fact, sCM can be stably scaled to modern text-to-image diffusion models and greatly improve the generation speed and 1-step generation quality!

Excited for 🏃SANA-Sprint. 🚀Code and weights will be released very soon along with diffusers. Study tuned!❤️

Introducing Sana-1.5. Model scaling up, then scaling down. Also inference time scaling is working as an auto end to end pipeline.