Image to LoRA Workflow

This workflow lets you turn a single reference image into a fully-trained, consistent AI character using FLUX.1 and ComfyUI.

1. Prompt Generation
Gemini analyzes your reference image and generates 20 diverse text prompts using advanced LLM inference.

2. Image Generation
FLUX.1 Kontext model takes those prompts and produces 20 consistent visual variations of your image.

3. Dataset Creation
All generated images are automatically saved, each with an associated caption file—ready for training.

4. LoRA Training
Train your custom LoRA directly within ComfyUI using the integrated pipeline.

Custom Nodes

via ComfyUI Manager:

• ComfyUI-FluxTrainer

• ComfyUI-SP-Nodes

• ComfyUI-Custom-Scripts

• WAS Node Suite

• ComfyUI_Fill-Nodes

• ComfyUI-Impact-Pack

• ComfyUI-KJNodes

• comfyui_tinyterranodes

Models

• Main Model: FLUX.1 Kontext (flux1-dev-kontext_fp8_scaled.safetensors)

• VAE: FLUX VAE (ae.safetensors)

• Encoders: FLUX CLIP (clip_l.safetensors, t5xxl_fp8_e4m3fn_scaled.safetensors)

• Training Models: flux1-dev.safetensors, t5xxl_fp16.safetensors

Setup Instructions

1. Load the workflow.json file in ComfyUI

2. Upload your reference image

3. Input your Gemini API key

4. Configure the following parameters:

   • Trigger Word: Unique identifier for your LoRA

   • Folder Name: Directory to save your dataset and LoRA

   • Output Path: ComfyUI output directory

5. Run the Workflow

Your trained LoRA will be saved automatically in the folder you specified.

Created by Levis Odin