Skip to content

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.

Fix k_diffusion math being off by a tiny bit during txt2img.

comfyanonymous036a22077ca3
diff --git a/comfy/samplers.py b/comfy/samplers.py
index dffd7fe7c..f83b2095b 100644
--- a/comfy/samplers.py
+++ b/comfy/samplers.py
@@ -650,7 +650,10 @@ class KSampler:
self.model_k.latent_image = latent_image
self.model_k.noise = noise
- noise = noise * sigmas[0]
+ if max_denoise:
+ noise = noise * torch.sqrt(1.0 + sigmas[0] ** 2.0)
+ else:
+ noise = noise * sigmas[0]
k_callback = None
total_steps = len(sigmas) - 1