NobodyButMeow

Allows more realistic images by using Guidance Scale of 2.2 instead of the default 3.5Adapted from https://openart.ai/workflows/-/-/3EhdlEU8Edh3xgjYsIQDYou can change the guidance scale (default 2.2) by using the complete workflow: https://tensor.art/workflows/761603486893821192

This tool is made so that you can enter different prompts for the clip_l, clip_g, and t5 text encoders.The reason for doing that is for prompt enhancement and other "hacks". See this video by Matt3o, the creator of the IPAdapter ComfyNode about how to work around some of SD3 Medium's problems: https://www.youtube.com/watch?v=OrST6Nq1NUg (the good stuff starts at around 8:50) TLDR: Various ways to "hack" SD3 when you see bad generation: Use non-standard resolution such as 1042x1042 (not multiple of 64).Avoid words such as "lying". Or use the 2 CLIP + T5 ClipTextEncoderSD3 node, but only use "lying" in clip_g.Use "random noise" such as "aaaaa aaaaaaa aaaaa" in your negative prompt.