ComfyUI_examples
Nvidia Cosmos Models
Nvidia Cosmos is a family of “World Models”. ComfyUI currently supports specifically the 7B and 14B text to video diffusion models and the 7B and 14B image to video diffusion models.
Files to Download
You will first need:
Text encoder and VAE:
oldt5_xxl_fp8_e4m3fn_scaled.safetensors goes in: ComfyUI/models/text_encoders/
cosmos_cv8x8x8_1.0.safetensors goes in: ComfyUI/models/vae/
Note: oldt5_xxl is not the same as the t5xxl used in flux and other models. oldt5_xxl is t5xxl 1.0 while the one used in flux and others is t5xxl 1.1
Video Models
The video models can be found in safetensors format here.
The workflows on this page use Cosmos-1_0-Diffusion-7B-Text2World.safetensors and Cosmos-1_0-Diffusion-7B-Video2World.safetensors
These files go in: ComfyUI/models/diffusion_models
Note: “Text to World” means Text to video and “Video to World” means image/video to video.
If you want the original diffusion models in .pt format instead of the repacked safetensors the official links are: 7B-Text2World 7B-Video2World 14B-Text2World 14B-Video2World
Workflows
Text to Video
This workflow requires the 7B text to video model that you can download above.
Image to Video
This model supports generating a video from 1 or more images. If more than one image is fed it will use them all as a guide and continue the motion. You can also do basic interpolation by setting one or more start_image and end_image which works best if those images are similar to each other.
This workflow requires the 7B image to video model that you can download above.
This model is trained primarily on realistic videos but in this example you can see that it also works decently on anime.
