How to Train LoRA in ComfyUI Using Flux - Full Step-by-Step Tutorial
Vladimir Chopine [GeekatPlay] via YouTube
Overview
Syllabus
00:00 - Introduction to LoRA training in ComfyUI
00:17 - Downloading and setting up ComfyUI 0.36
00:42 - Extracting and preparing folders
00:49 - Installing ComfyUI Manager
01:28 - Opening terminal and cloning ComfyUI Manager
02:11 - Launching ComfyUI with GPU batch file
02:36 - Checking if Manager tab is available
02:54 - Installing Flux Trainer nodes
03:19 - Installing RGthree nodes
03:48 - Installing KJNodes
03:59 - Installing WAS Node Suite optional
04:22 - Restarting ComfyUI after node installation
04:44 - Optional: using BLIP to generate captions
05:56 - BLIP captioning workflow setup
07:00 - Downloading required Flux server files
07:50 - Placing models and encoders in correct folders
08:11 - Downloading and loading the training workflow JSON
08:46 - Fixing missing nodes if needed
09:04 - Exploring the loaded workflow structure
09:56 - Preparing dataset and verifying paths
10:18 - Connecting models and adjusting parameters
11:01 - Running LoRA training in ComfyUI
12:18 - Training time overview and system specs
13:01 - Previewing training loss and performance
14:44 - Reviewing output files and trained models
15:50 - Testing the trained LoRA model
16:45 - Comparing results with and without LoRA
17:21 - Final thoughts and training tips
Taught by
Vladimir Chopine [GeekatPlay]