Flux LoRA Dataset Composer + UpScaler

“Flux LoRA Dataset Composer + UpScaler”
Pee Nueng, Phuwadol Mitthanond, explains in detail how to prepare a LoRA dataset for your face.

• Starting from the basic working principles
• How to prepare and select the appropriate images
• Using the Dataset Composer tool, which can be accessed via phone, tablet, or computer
• Image cropping
• Generating captions with varo-ai
• Upscaling (Pee Nueng mentions that it’s essential), which was recently updated to improve the resolution of small images used for training. The Upscale feature in Dataset Composer does not use SD Upscale or Multidiffusion Upscale but instead utilizes Real-ESRGAN+GFPGAN. These tools are commonly used for image restoration and enlargement. The benefit is that the images become sharper, more detailed, fit the system’s requirements perfectly, and the faces do not change. It’s especially suitable for preparing datasets for training individual LoRA models. ☺️
• Training in Auto mode
• And testing the image generation for you to see

This video explains the training process for Flux LoRA to achieve high-quality individual LoRA models with the most straightforward steps ever.
Feel free to bookmark this clip! 🥰