AI Image Generation with Stable Diffusion Part 2 | Img2Img Transformations, Masking, Upscaling
Part 2 of a look at image generation with Stable Diffusion. Part 1 is up here: https://youtu.be/8xAq0-68jiU
I look at image to image transformations, masking, a little basic inpainting and outpainting, and installing the Python version of Real-ESRGAN with GFPGAN in Conda under Windows 10. For reference, the GPU used is an NVIDIA 1660 Super with 6gb. The amount of GPU memory limits the builds and features that can be used, and the card requires workarounds for upscaling.
Command for Real-ESRGAN +for GFPGAN for those that don't read docs (and 1660 users. Needs fp32 argument.):
python inference_realesrgan.py -n RealESRGAN_x4plus -i inputs --face_enhance --fp32
Installing Stable Diffusion and the Automatic1111 WebUI using Conda on Windows 10:
https://youtu.be/m_Bx1tTC8eE
Resources:
Real-ESRGAN Github Repo: https://github.com/xinntao/Real-ESRGAN
Models are available under the resources link, in the Assets section of each post.
Other upscalers using Real-ESRGAN:
https://github.com/WGzeyu/Real-ESRGAN-EGUI
https://github.com/net2cn/Real-ESRGAN_GUI
https://github.com/AaronFeng753/Waifu2x-Extension-GUI
https://github.com/Justin62628/Squirrel-RIFE
https://github.com/shangar21/anime_upscaler
Miniconda for Windows: https://docs.conda.io/en/latest/miniconda.html
Stable Diffusion - HLKY Fork [with documentation]: https://github.com/hlky/stable-diffusion
Stable Diffusion official repo: https://github.com/CompVis/stable-diffusion
GPU-Z download: https://www.techpowerup.com/gpuz/
Real-ESRGAN (Covering in Part 2): https://github.com/xinntao/Real-ESRGAN
Stable Diffusion ultimate GUI install guide for Windows: https://rentry.org/GUItard
Stable Diffusion Linux install guide with scripts: https://github.com/hlky/stable-diffusion/wiki/Linux-Installation
CFG explained, but I'm still lost: https://benanne.github.io/2022/05/26/guidance.html
Stable Diffusion CPU guide: https://rentry.org/cputard
Lexica Art / Prompt Search: https://lexica.art/