This Chroma workflow allows you to:

1. Load an image from Civitai or locally

2. If the image has a prompt or tags, it uses this as a starter prompt to feed into Ollama Locally

3. Either way, Ollama will refine the prompt based on image and the metadata into a descriptive natural language prompt, including NSFW components

4. Lora stack added

5. The image is also run through controlnet with customizable settings

6. After the first sampling is complete, the image is then run through 5 Detailers, you can toggle these or change the Bbox model they are using for detection, for now it details:

   1. Face

   2. Vagina

   3. Breasts

   4. Penis

   5. Overall

7. From here, the image is put through a slight fine grain addition to be saved.

To edit a detailer, click the subgraph expansion icon on the top right of each detailer, so you can make modifications like changing the detection model, prompt for that detailer etc.

You can swap out the LLM for a different node if you like, such as Gemini API

You can replace words it finds in the original image prompt BEFORE it goes through LLM prompting but also AFTER in the final LLM prompt you can also find/replace words. Eg Blonde to Ginger.

Custom Nodes that are active:

• ComfyUI Image & Prompt Loader

• ComfyUI-Custom-Scripts

• ComfyUI Impact Pack

• ComfyUI_LayerStyle

• rgthree-comfy

• ComfyUI-KJNodes

• ComfyUI_essentials

• CG Use Everywhere

• RES4LYF

• ComfyUI Impact Subpack

• comfyui-propost

• ComfyUI-OllamaGemini

• WAS Node Suite (Revised)

• ComfyUI-LogicUtils

• ComfyLiterals