The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
Add noise augmentation to hunyuan image refiner. (#9831)
comfyanonymousfd2b820ec28e
diff --git a/comfy/model_base.py b/comfy/model_base.pyindex c69a9d1ad..8422051bf 100644--- a/comfy/model_base.py+++ b/comfy/model_base.py@@ -1437,6 +1437,7 @@ class HunyuanImage21Refiner(HunyuanImage21):def concat_cond(self, **kwargs):noise = kwargs.get("noise", None)image = kwargs.get("concat_latent_image", None)+ noise_augmentation = kwargs.get("noise_augmentation", 0.0)device = kwargs["device"]if image is None:@@ -1446,6 +1447,9 @@ class HunyuanImage21Refiner(HunyuanImage21):image = utils.common_upscale(image.to(device), noise.shape[-1], noise.shape[-2], "bilinear", "center")image = self.process_latent_in(image)image = utils.resize_to_batch_size(image, noise.shape[0])+ if noise_augmentation > 0:+ noise = torch.randn(image.shape, generator=torch.manual_seed(kwargs.get("seed", 0) - 10), dtype=image.dtype, device="cpu").to(image.device)+ image = noise_augmentation * noise + (1.0 - noise_augmentation) * imagereturn imagedef extra_conds(self, **kwargs):diff --git a/comfy_extras/nodes_hunyuan.py b/comfy_extras/nodes_hunyuan.pyindex 351a7e2cb..db398cdf1 100644--- a/comfy_extras/nodes_hunyuan.py+++ b/comfy_extras/nodes_hunyuan.py@@ -134,6 +134,7 @@ class HunyuanRefinerLatent:return {"required": {"positive": ("CONDITIONING", ),"negative": ("CONDITIONING", ),"latent": ("LATENT", ),+ "noise_augmentation": ("FLOAT", {"default": 0.10, "min": 0.0, "max": 1.0, "step": 0.01}),}}RETURN_TYPES = ("CONDITIONING", "CONDITIONING", "LATENT")@@ -141,11 +142,10 @@ class HunyuanRefinerLatent:FUNCTION = "execute"- def execute(self, positive, negative, latent):+ def execute(self, positive, negative, latent, noise_augmentation):latent = latent["samples"]-- positive = node_helpers.conditioning_set_values(positive, {"concat_latent_image": latent})- negative = node_helpers.conditioning_set_values(negative, {"concat_latent_image": latent})+ positive = node_helpers.conditioning_set_values(positive, {"concat_latent_image": latent, "noise_augmentation": noise_augmentation})+ negative = node_helpers.conditioning_set_values(negative, {"concat_latent_image": latent, "noise_augmentation": noise_augmentation})out_latent = {}out_latent["samples"] = torch.zeros([latent.shape[0], 32, latent.shape[-3], latent.shape[-2], latent.shape[-1]], device=comfy.model_management.intermediate_device())return (positive, negative, out_latent)