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

Fix "Load Checkpoint with config" node.
comfyanonymous committed 3 years ago
fe4c074
Use the GPU for the canny preprocessor when available.
comfyanonymous committed 3 years ago
d70b0bc
81d9200
Support SDXL t2i adapters with 3 channel input.
comfyanonymous committed 3 years ago
f2f5e5d
Remove optimization that caused border.
comfyanonymous committed 3 years ago
bed116a
check for text attr and save
Chris committed 3 years ago
18379de
copy metadata into modified image
Chris committed 3 years ago
edcff9a
feat(ui): Add node mode toggles to group context menu
Michael Poutre committed 3 years ago
e30d546
fbf375f
Put clip vision outputs on the CPU.
comfyanonymous committed 3 years ago
4e89b2c
Load clipvision model to GPU for faster performance.
comfyanonymous committed 3 years ago
a094b45
Move ModelPatcher to model_patcher.py
comfyanonymous committed 3 years ago
f92074b
Reduce min tile size for encode
BVH committed 3 years ago
d86b222
Implement loras with norm keys.
comfyanonymous committed 3 years ago
4798cf5
support on prompt event handler (#765)
Dr.Lt.Data committed 3 years ago
0faee11
Enable bf16-vae by default on ampere and up.
comfyanonymous committed 3 years ago
b8c7c77
Make --bf16-vae work on torch 2.0
comfyanonymous committed 3 years ago
d935ba5
fix: cannot disable dynamicPrompts (#1327)
Dr.Lt.Data committed 3 years ago
d9f4922
Fix lowvram model merging.
comfyanonymous committed 3 years ago
a57b0c7