🚀 Basic Workflow Tutorial

Learn to use the basic Kontext LoRA workflow for automatic character training

What You'll Learn

  • How to download and install the basic workflow
  • Setting up your first character training
  • Understanding automatic parameter selection
  • Monitoring training progress
  • Exporting your trained LoRA model

📥 Download Required Files

First, download the basic workflow JSON file:

Download Basic Workflow

Prerequisites

  • ComfyUI installed and working
  • FLUX.1 Kontext [dev] model downloaded
  • GPU with 8GB+ VRAM (recommended)
  • At least 16GB system RAM

Step-by-Step Instructions

Install Required Custom Nodes

Open ComfyUI-Manager and install these essential nodes:

# Required custom nodes: - ComfyUI-Manager - rgthree-comfy - ComfyUI-KJNodes - efficiency-nodes-comfyui

Download the Workflow

Download the basic-lora-trainer.json file and save it to your computer.

Import into ComfyUI

Drag and drop the JSON file into your ComfyUI interface. The workflow will automatically load with all nodes connected.

Prepare Your Reference Image

  • Use a high-quality image (1024x1024 recommended)
  • Character should take up most of the frame
  • Clear lighting and minimal background clutter
  • JPG, JPEG, or PNG format

Configure the Workflow

The basic workflow uses automatic settings, but you can adjust:

  • Character name: Enter a unique name for your character
  • Training steps: Default 1000 (increase for higher quality)
  • Output path: Where to save your trained model

Start Training

Click "Queue Prompt" to begin training. The process typically takes 5-15 minutes depending on your GPU.

Training Progress: Step 100/1000 - Loss: 0.125 Step 200/1000 - Loss: 0.089 Step 500/1000 - Loss: 0.045 Step 1000/1000 - Complete!

Test Your Model

Once training completes, test your LoRA model:

  • Load the saved .safetensors file in ComfyUI
  • Try generating images with your character name
  • Experiment with different prompts and styles

Tips for Best Results

  • Quality images: Use the highest quality reference image possible
  • Unique names: Choose distinctive character names to avoid conflicts
  • Monitor VRAM: Keep an eye on memory usage during training
  • Patience: Let training complete fully for best quality

Next Steps

Once you've mastered the basic workflow, consider: