From 22aea7f31cf9702f2e800c88dad1407b2aac5f6e Mon Sep 17 00:00:00 2001 From: File_xor Date: Mon, 29 Jun 2026 03:47:30 +0900 Subject: [PATCH] add README.md --- README.md | 10 ++++++++++ 1 file changed, 10 insertions(+) create mode 100644 README.md diff --git a/README.md b/README.md new file mode 100644 index 0000000..77d9199 --- /dev/null +++ b/README.md @@ -0,0 +1,10 @@ +# SEGA Anima + +SEGA anima is implementation of ["SEGA: Spectral-Energy Guided Attention for Resolution Extrapolation in Diffusion Transformers" (arXiv:2605.22668)](https://arxiv.org/abs/2605.22668) for Anima model in Comfyui. + +## What should I do when output image is bad + +1. First thing you should to do is increasing CFG. If you could not find good CFG, come back later. +2. Reduce "training_resolution" if image does not appear. Too low "training_resolution" results structual inaccuracy. In that case, increase "training_resolution" or go to next step. +3. You may increase "theta" or "base_mscale_coefficient" or "d_mul". +4. Creator of the this node does not know what rest of parameters do. You may do experience with those. \ No newline at end of file