Using Stable Diffusion and DreamBooth to Create Meme Art

Learn how to create meme art using stable diffusion and DreamBooth in this comprehensive 36-minute tutorial. Explore the process of training on and generating variations of Final Fantasy XIV characters, then compositing them for a final meme piece. Follow along as the instructor demonstrates DreamBooth training, character masking, img2img techniques, and image refinement in 2D software. Gain insights into prompt engineering, checkpoint conversion, and useful add-ons for achieving painterly effects. Discover tips for correcting image irregularities, adding camera blur, and implementing atmospheric perspective. Perfect for digital artists, machine learning enthusiasts, and fans of AI-generated art looking to enhance their meme creation skills.

- Intro
- Final pieces
- Inspiration behind this experiment/workflow
- DreamBooth process in WSL2
- Confirming prompt token from new checkpoint in AUTOMATIC1111 webui
- Masking out character in photoshop
- Using masked out character as base initialization for warrior of light in img2img
- Same process for Venat
- Same process for Meteion
- Making changes to base meme masked character for better img2img
- Generating background image in img2img from a personal photograph
- Masking out generated character in photoshop
- Tips and tricks for correcting a base image
- How to get camera blur on background image
- Brief explanation of atmospheric/color perspective
- Adding atmospheric haze to characters pick a diffuse color, paint bucket a layer, put it on hard light, adjust opacity
- Conclusion