This is a simplified guide to an AI model called Any-Comfyui-Workflow maintained by Comfyui. If you like these kinds of analysis, you should join AImodels.fyi or follow us on Twitter.
Model overview
any-comfyui-workflow lets you run any ComfyUI workflow on Replicate’s infrastructure. Created by comfyui, this model accepts workflows in API format and handles the execution on GPU hardware. Similar alternatives include any-comfyui-workflow by fofr and any-comfyui-workflow-a100 for higher-end hardware deployments. The model includes many popular model weights and custom nodes out of the box, though you can request additional ones or bring your own.
Model inputs and outputs
any-comfyui-workflow accepts a ComfyUI workflow JSON file along with optional input files like images or archives. The model processes your workflow and returns generated output files. You can control output format, quality, seed behavior, and cache settings to customize each run.
Inputs
- workflow_json – Your ComfyUI workflow as a JSON string or URL. Must be in API format, obtained from ComfyUI using the “Save (API format)” button
- input_file – Optional input image, video, tar, or zip file. Alternatively, replace inputs with URLs in your JSON workflow and the model will download them
- output_format – Format for output images: webp, jpg, or png (default: webp)
- output_quality – Quality of output images from 0 to 100 (default: 95)
- randomise_seeds – Automatically randomize seeds like seed, noise_seed, and rand_seed (default: true)
- force_reset_cache – Force reset the ComfyUI cache before running the workflow for debugging
- return_temp_files – Return temporary files such as preprocessed controlnet images for debugging
Outputs
- Array of image URIs – URLs to the generated output images from your workflow
Capabilities
The model executes complete ComfyUI wo…