/Comfy-Topaz

Topaz Photo AI integration for ComfyUI

Primary LanguagePythonMIT LicenseMIT

What is Comfy-Topaz?

Comfy-Topaz is a custom node for ComfyUI, which integrates with Topaz Photo AI to enhance (upscale, sharpen, denoise, etc.) images, allowing this traditionally asynchronous step to become a part of ComfyUI workflows.

Requirements

  • Licensed installation of Topaz Photo AI: This provides tpai.exe, the path to which should be set for the Topaz Photo AI (tpai.exe) setting in ComfyUI.

Installation

Clone this repo into ComfyUI/custom_nodes and restart ComfyUI.

Usage

Auto-Pilot Settings

This is the simplest use case, which relies on Topaz Photo AI to auto-detect and apply those settings. This is done by omitting the upscale and sharpen settings as inputs:

simple demo showing the auto-pilot settings

Sometimes auto-pilot settings don't yield the best results, which warrants manual tuning.

On the output side, autopilot_settings shows what the auto-pilot settings were, and settings shows all the features used and knobs turned to generate the final image.

Manual Settings

Override auto-pilot settings by providing manual settings:

demo showing the upscale settings override

A good starting point is by copying over params from auto-pilot from which to iterate. I copied over the denoise, deblur, and detail values and changed the model from Standard V2 to High Fidelity.

TODO

  • Output *Settings nodes rather than json, to eliminate the manual steps of copying values when overriding settings.
  • Add a button to run auto-pilot analysis without applying the settings on the image.
  • Map param1, param2, param3, ... to the actual param name (e.g., denoise, deblur, and detail for Upscale Settings)
  • Expose more settings (denoise, face recovery, text recovery, WB/exposure adjustment).