Comfy-agent workflow_ep51_cosmos_predict2_2b_text2image

name: workflow_ep51_cosmos_predict2_2b_text2image description: Workflow wrapper for EP51 Cosmos Predict2 2B Text2Image.json inputs: prompt: type: string required: false negative_prompt: type: string required: false width: type: integer required: false height: type: integer required: false seed: type: integer required: false steps: type: integer required: false cfg: type: number required: false sampler_name: type: string required: false scheduler: type: string required: false denoise: type: number required: false server: type: string required: false headers: type: object required: false api_prefix: type: string required: false outputs: status: type: string prompt_id: type: string output_images: type: array requirements: models:

• type: vae name: wan_2.1_vae.safetensors target_folder: models/vae
• type: clip name: oldt5_xxl_fp8_e4m3fn_scaled.safetensors target_folder: models/clip
• type: diffusion_model name: cosmos_predict2_2B_t2i.safetensors target_folder: models/diffusion_models custom_nodes: [] links:
• https://discord.com/invite/gggpkVgBf3
• https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/resolve/main/cosmos_predict2_2B_t2i.safetensors?download=true
• https://huggingface.co/Comfy-Org/Cosmos_Predict2_repackaged/tree/main
• https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors
• https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/resolve/main/text_encoders/oldt5_xxl_fp8_e4m3fn_scaled.safetensors
• https://www.youtube.com/@pixaroma input_modalities:
• text_prompt output_modalities:
• image/png model_families:
• sd3
• wan node_count: 10 node_types:
• CLIPLoader
• CLIPTextEncode
• EmptySD3LatentImage
• KSampler
• MarkdownNote
• SaveImage
• UNETLoader
• VAEDecode
• VAELoader selection_metadata: family: txt2img resource_profile: medium complexity_score: 4 estimated_runtime: moderate (about 30-120s depending on server) warnings:
• Large model(s) detected; ensure enough VRAM and disk space.
• Medium-high resolution detected; expect moderate extra runtime. max_width: 1024 max_height: 1024 max_steps: 35