xinyun/ComfyUI-KJNodes
1
0
Fork 0
mirror of https://git.datalinker.icu/kijai/ComfyUI-KJNodes.git synced 2025-12-09 21:04:41 +08:00
Code Issues Packages Projects Releases Wiki Activity

Update nodes.py

Browse Source
This commit is contained in:
kijai 2024-03-23 16:46:59 +02:00
parent f9563afd7b
commit 3015008b1c
1 changed files with 4 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
nodes.py
View File

@ -2826,6 +2826,7 @@ class InjectNoiseToLatent:
"optional":{ "optional":{
"mask": ("MASK", ), "mask": ("MASK", ),
"mix_randn_amount": ("FLOAT", {"default": 0.0, "min": 0.0, "max": 1000.0, "step": 0.001}), "mix_randn_amount": ("FLOAT", {"default": 0.0, "min": 0.0, "max": 1000.0, "step": 0.001}),
"seed": ("INT", {"default": 123,"min": 0, "max": 0xffffffffffffffff, "step": 1}),
} }
} }
@ -2834,7 +2835,7 @@ class InjectNoiseToLatent:
CATEGORY = "KJNodes/noise" CATEGORY = "KJNodes/noise"
def injectnoise(self, latents, strength, noise, normalize, average, mix_randn_amount=0, mask=None): def injectnoise(self, latents, strength, noise, normalize, average, mix_randn_amount=0, seed=None, mask=None):
samples = latents.copy() samples = latents.copy()
if latents["samples"].shape != noise["samples"].shape: if latents["samples"].shape != noise["samples"].shape:
raise ValueError("InjectNoiseToLatent: Latent and noise must have the same shape") raise ValueError("InjectNoiseToLatent: Latent and noise must have the same shape")
@ -2851,6 +2852,8 @@ class InjectNoiseToLatent:
mask = mask.repeat((noised.shape[0] -1) // mask.shape[0] + 1, 1, 1, 1)[:noised.shape[0]] mask = mask.repeat((noised.shape[0] -1) // mask.shape[0] + 1, 1, 1, 1)[:noised.shape[0]]
noised = mask * noised + (1-mask) * latents["samples"] noised = mask * noised + (1-mask) * latents["samples"]
if mix_randn_amount > 0: if mix_randn_amount > 0:
if seed is not None:
torch.manual_seed(seed)
rand_noise = torch.randn_like(noised) rand_noise = torch.randn_like(noised)
noised = ((1 - mix_randn_amount) * noised + mix_randn_amount * noised = ((1 - mix_randn_amount) * noised + mix_randn_amount *
rand_noise) / ((mix_randn_amount**2 + (1-mix_randn_amount)**2) ** 0.5) rand_noise) / ((mix_randn_amount**2 + (1-mix_randn_amount)**2) ** 0.5)