Stable Diffusion - Create a PNGtuber with Character LoRA

This course demonstrates a complete Stable Diffusion workflow for creating a PNGtuber, including character LoHa training, style LoRA prompting, image compositing, expression preparation, and setup in veadotube mini with OBS.

Syllabus

Intro
Tools you'll need
Dataset introduction
Reasoning for using a character LoHa
Captioning the character
Character LoHa training settings
Initial test of character LoHa
Choosing the best epoch
Verifying the chosen epoch LoHa integrity
Mini veadotube mini overview
Choosing the style LoRA
Check the raw style LoRA
Prompting with both character LoHa and style LoRA
Technique to vary seed within a radius
Compositing bust shots together
Overview of base full body image
Outpainting the full body
Outpainting the bust shot
Upscale the body
Upscale the bust
Separate head from background
Preparing the body for bust photobash
Rasterizing a smart object
Liquifying for better seams
Refining the entire character
Preparing head for smart object simplifies character expression replacement
Creating the open mouth expression
Creating the closed mouth expression
Creating the closed eyes expression
Making the closed eyes a global toggle
Removing the background from the entire character
Exporting in the correct format for veadotube mini
Setting up pngtuber in veadotube mini
Using veadotube mini with OBS
Closing thoughts