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.
- [1.1](./wyrde branching shapren x6-v1.1.json) Cleaned some noodles. Altered some denoise settings. Using the configuralbe save file prefix.
WAS nodes. https://github.com/WASasquatch/was-node-suite-comfyui
Model
Lora
- none this time, but can be added easily
Uspcale
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