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.

Commits on master

f81a6fa
Don't load the model in SDTurboScheduler
comfyanonymous committed 2 years ago
94d5a12
Workaround xformers bug.
comfyanonymous committed 2 years ago
2aed53c
Add TLS Support (#3312)
Garrett Sutula committed 2 years ago
bacce52
Update Readme.
comfyanonymous committed 2 years ago
bb8b48a
Add a SamplerLCMUpscale node.
comfyanonymous committed 2 years ago
eecd69b
059773a
Add a denoise value to AlignYourStepsScheduler.
comfyanonymous committed 2 years ago
10fcd09
8cab3be
7990ae1
Free more vram for upscale models.
comfyanonymous committed 2 years ago
16eabdf
Don't init a VAE model when there are no VAE weights.
comfyanonymous committed 2 years ago
8dc19e4
Increase max lora strength to 100.0
comfyanonymous committed 2 years ago
27d5808
Free some memory before loading upscale models.
comfyanonymous committed 2 years ago
d09b5ef
Speed up Sharpen node.
comfyanonymous committed 2 years ago
4ee9aad
644a3ae
Faster blur.
comfyanonymous committed 2 years ago
133dc33
Sync some minor changes from the other repo.
comfyanonymous committed 2 years ago
5d08802
Support VAE without quant_conv.
comfyanonymous committed 2 years ago
c59fe9f
Improve node input/widget conversion sub-menus (#3281)
Torbjörn Lönnemark committed 2 years ago
a88b0eb
abc69ca
Implement PerpNeg as a guider.
comfyanonymous committed 2 years ago
45ec1cb
This can be removed since PAG doesn't use the uncond.
comfyanonymous committed 2 years ago
8903dce
Add basic PAG node.
comfyanonymous committed 2 years ago
719fb2c
Fix some memory related issues.
comfyanonymous committed 2 years ago
258dbc0
744ac94
Support SDXS 512 model.
comfyanonymous committed 2 years ago
58812ab