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

Ignore origin domain when it's empty.
comfyanonymous committed 2 years ago
e3b0402
fix: url decode filename from API (#4801)
Darion committed 2 years ago
967867d
Fix issue with last commit.
comfyanonymous committed 2 years ago
cbaac71
Fix lora issue.
comfyanonymous committed 2 years ago
9c5fca7
Support onetrainer text encoder Flux lora.
comfyanonymous committed 2 years ago
32a60a7
Fix import issue (#4815)
Jim Winkens committed 2 years ago
bb52934
SaveLora node can now save "full diff" lora format.
comfyanonymous committed 2 years ago
8aabd7c
a09b29c
9bfee68
Support a generic Comfy format for text encoder loras.
comfyanonymous committed 2 years ago
ea77750
Fix onnx export not working on flux.
comfyanonymous committed 2 years ago
c27ebeb
dc2eb75
Update frontend to v1.2.47 (#4798)
Chenlei Hu committed 2 years ago
fa34efe
Mistoline flux controlnet support.
comfyanonymous committed 2 years ago
5cbaa9e
Prioritize freeing partially offloaded models first.
comfyanonymous committed 2 years ago
c742737
22d1241
Add emb_patch support to UNetModel forward (#4779)
Jedrzej Kosinski committed 2 years ago
f04229b
Support newer glora format.
comfyanonymous committed 2 years ago
483004d
Lower fp8 lora memory usage.
comfyanonymous committed 2 years ago
00a5d08
Flux onetrainer lora.
comfyanonymous committed 2 years ago
d043997
fix typo in stale-issues (#4735)
Alex "mcmonkey" Goodwin committed 2 years ago
f1c2301
Speed up inference on nvidia 10 series on Linux.
comfyanonymous committed 2 years ago
8d31a66
b643eae
Update manual install instructions.
comfyanonymous committed 2 years ago
baa6b4d
d4aeefc
Remove github buttons.
comfyanonymous committed 2 years ago
587e7ca