---

license: other license_name: nvidia-open-model-license license_link: >- https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/ base_model:

• nvidia/Cosmos-Predict2-2B-Text2Image

pipeline_tag: text-to-image tags:

• gguf-node

widget:

• text: >-

On a rainy night, a girl holds an umbrella and looks at the camera. The rain keeps falling. output: url: samples\Cosmos00001_.png

• text: >-

On a rainy night, a girl holds an umbrella and looks at the camera. The rain keeps falling. a serene atmosphere output: url: samples\Cosmos00002_.png

• text: >-

On a rainy night, a girl holds an umbrella and looks at the camera. The rain keeps falling. output: url: samples\Cosmos00003_.png

• text: >-

On a rainy night, a girl holds an umbrella and looks at the camera. The rain keeps falling. output: url: samples\Cosmos00004_.png

• text: >-

The night of the city, from the perspective of the car's front window, real shooting, real scene, late-night emo style, emotional, sad, neon lights, sense of the future, rain, cinematic feel, cyberpunk output: url: samples\Cosmos00005_.mp4

• text: >-

The night of the city, from the perspective of the car's front window, real shooting, real scene, late-night emo style, emotional, sad, neon lights, sense of the future, rain, cinematic feel, cyberpunk output: url: samples\Cosmos00006_.mp4

---

gguf quantized version of cosmos-predict2 (incl. 2b and 14b)

• drag cosmos to > ./ComfyUI/models/diffusionmodels
• drag t5xxl-old to > ./ComfyUI/models/textencoders
• drag pig to > ./ComfyUI/models/vae

!screenshot

!screenshot

• don't need safetensors anymore; all gguf (model + encoder + vae)
• full set gguf works on gguf-node (see the last item from reference below)
• get more t5xxl-old gguf encoder either here or here
• note (for internal use) - ex: extra state excluded; ex3: ex + posembedder x 3 excluded

reference

• base model from nvidia
• comfyui from comfyanonymous
• gguf-node (pypi|repo|pack)