SANA Video 2.0
SANA Video 2.0 is a hybrid video diffusion transformer built at 5B and 14B parameter scales that generates video up to 720p resolution on a single GPU. It combines gated linear attention with periodic softmax attention anchors in a 3 to 1 ratio to reduce computation while preserving quality comparable to full softmax video models, and uses Block Attention Residuals to propagate refined features across network depth for efficient long sequence video generation.
- Type
- Video
- Weights
- Open
- Downloads
- 1.1k
- Likes
- 11
- Pipeline
- text to video
- License
- apache-2.0
Model hub
Efficient-Large-Model/SANA-Video_2.0_5B_720p
- Repo
- Efficient-Large-Model/SANA-Video_2.0_5B_720p
- Library
- sana
- Access
- Public
Technical tags and languages
More from NVIDIA
11
-
39k DL
Alpamayo 1.5 10B
Multimodal
Alpamayo 1.5-10B is NVIDIA’s open 10B vision-language-action model for autonomous driving. It is built as a s…
-
12k DL
Alpamayo R1
Text
Alpamayo-R1 is NVIDIA's vision-language-action model for autonomous driving, combining Chain of Causation rea…
-
4.1k DL
Asset Harvester
Multimodal
Asset Harvester is NVIDIA’s open framework for extracting reusable 3D assets from sparse autonomous-driving o…
-
1.3M DL
bigvgan_v2_22khz_80band_256x
Audio
audio to audio · License mit · 1,327,174 downloads · by nvidia
-
829k DL
bigvgan_v2_44khz_128band_512x
Audio
audio to audio · License mit · 828,966 downloads · by nvidia
-
3k DL
Canary 180m flash
Audio
Canary 180M Flash is a compact automatic speech recognition model in NVIDIA’s Canary family, designed for fas…
-
39k DL
Canary 1B v2
Audio
Multilingual, multitask speech recognition model (ASR) in the NVIDIA NeMo family.
-
98 DL
ChronoEdit 14B Diffusers Upscaler LoRA
Image
ChronoEdit-14B is a diffusion-based image and video editor with temporal consistency and built-in upscaling.…
About NVIDIA
1