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.

Tweak hunyuan memory usage factor.

comfyanonymous88ceb28e2094
diff --git a/comfy/supported_models.py b/comfy/supported_models.py
index 87fecde5a..ff0bea418 100644
--- a/comfy/supported_models.py
+++ b/comfy/supported_models.py
@@ -788,7 +788,7 @@ class HunyuanVideo(supported_models_base.BASE):
unet_extra_config = {}
latent_format = latent_formats.HunyuanVideo
- memory_usage_factor = 1.7 #TODO
+ memory_usage_factor = 1.8 #TODO
supported_inference_dtypes = [torch.bfloat16, torch.float32]