# wyrde branch sharpen HR-Fix model upscaling x4 (many images)
Takes an image from text prompt (or image) and
* samples it
* uses an upscale model on it
* reduces it again and sends to a pair of samplers
* they upscale and reduce
* then send results to another pair of samplers
Not quite a true double-pass sharpen, but the results are decent. Lots of variation.
Unhook noodles to reduce undesired images and paths. Move noodles to switch VAE and img2img.
## example
1st image
Images 2A and 2B 
Images 3A - 3D 


## resources
WAS nodes. https://github.com/WASasquatch/was-node-suite-comfyui
Model
* https://civitai.com/models/8281/perfect-world
Lora
* none this time, but can be added easily
Uspcale
* Lollypop https://drive.google.com/u/1/uc?id=10h8YXKKOQ61ANnwLjjHqXJdn4SbBuUku&export=download
Embeds
* EasyNegative https://civitai.com/models/7808/easynegative
* bad-hands-5 https://huggingface.co/yesyeahvh/bad-hands-5/tree/main
* bad-artist: https://huggingface.co/nick-x-hacker/bad-artist