I can't get it to work. Tried with realistic vision v1.3. The images are still bright and almost the same as without lora.

s this really good with anime model?

Great LORA!! One question -- this doesn't seem to have any effect when inpainting (the image becomes very flat, losing all the lighting texture), is that expected?

i keep getting an error when trying to use this ... v2 doesnt work at all .. v2 pynoise does load but gives me runtime errors ?

When using this LoRA in inpainting mode, it seems to make the image brighter/lighter rather than darker. Increasing the weight makes it even brighter. Anyone else notice this?

Please, fill the Model Info for v2.

Is the a model for Pyramid noise u extracted the LoRA from? if so please link me to it.

Cannot find: "lora_te_wrapped_transformer_text_model_encoder_layers_11_self_attn_q_proj", skipped

Cannot find: "lora_te_wrapped_transformer_text_model_encoder_layers_11_self_attn_out_proj", skipped

Cannot find: "lora_te_wrapped_transformer_text_model_encoder_layers_11_mlp_fc1", skipped

Cannot find: "lora_te_wrapped_transformer_text_model_encoder_layers_11_mlp_fc2", skipped

what happened??

anyone managed to use it with Draw Things mac app please? Struggling to import it, any help would be greatly appreciated

I am using additional networks on automatic 1111 and the about section stated to increase the limit to go over two for the pynoise. I can not find this config file and I edited the additional_netowrks.py and I can not get it to go higher than two. Help please ;-;

Please, please, make a SD 2.1 Version. This Lora ist sooooo good.

excellent. to many images come out bright. no matter what. gotta pop things into adobe or get creative with prompts

"If u use Additional Networks u have to push the Max Limit in the config file over 2 to get more depth."

Can you explain this a little more? In the config there are a few settings for Additional Networks, but none for Max Limit. I tried adding a Max Limit setting, but it had no effect. Thanks.

Is it possible to change the "two tone lighting" tag to something different?

The issue is that if you add stronger modifiers to it, it makes either hair or other subjects in the picture change to two-tone as well. I'm not well-versed in LoRA, so not sure if this is possible.

Not able to download the safetensor file. Could someone tell me how do i get to download this? thanks in advance

The core reason for too bright images was just found:

https://www.reddit.com/r/StableDiffusion/comments/13joe98/sds_noise_schedule_is_flawed_this_new_paper/

Help with weights? Using a higher weight on this vs other things means the other weighs are... less weighted. Some generations looked great with <lora:epiNoiseoffset_v2-pynoise:6>, but my other prompts got so much less attention. Solution?

Copy the epiNoiseoffset_v2-pynoise.safetensors file multiple times. I did this 6 times(add _1 _2 _3....) Added each of those 6 to the prompt giving a weight of 1. This doesn't exactly give you <lora:epiNoiseoffset_v2-pynoise:6>, but it heads in that direction.

I might also be dumb, putting the same reference in there multiple times to the same file might do the same...

<lora:epiNoiseoffset_v2-pynoise:1>

<lora:epiNoiseoffset_v2-pynoise:1>

<lora:epiNoiseoffset_v2-pynoise:1>

<lora:epiNoiseoffset_v2-pynoise:1>

<lora:epiNoiseoffset_v2-pynoise:1>

<lora:epiNoiseoffset_v2-pynoise:1>

vs separate files looking like:

<lora:epiNoiseoffset_v2-pynoise_1:1>

<lora:epiNoiseoffset_v2-pynoise_2:1>

<lora:epiNoiseoffset_v2-pynoise_3:1>

<lora:epiNoiseoffset_v2-pynoise_4:1>

<lora:epiNoiseoffset_v2-pynoise_5:1>

<lora:epiNoiseoffset_v2-pynoise_6:1>

I don't understand the difference between this an LowRA, can anyone explain?

i can not see which basic model does every lora use, is that angthing? deliberate? realmix? or ????

Doesn't seem to work for me.
All it says is just ''Couldn't find network with name epiNoiseoffset_v2''

Hello, i cannot seem to get it to work. I add <lora:epi_noiseoffset2:1> and have placed the file inside stable-diffusion-webui/models/lora

Do i need to call it something else inside automatic1111?

This LoRA has the very nice side effect of increasing detail/coherence in my images (in comfyUI). It really polishes everything up when my images are lacking that last 15-20% 🪄

bruh this is the best and most useful lora i've ever found in this website

SDXL Version needed!
Great Work!

Exception: Couldn't identify epi_noiseoffset2.pt as neither textual inversion embedding nor diffuser concept.

Its not finding the lora <lora:epiNoiseoffset_v2:1> it is in my lora folder and still not recognizing it. I am using easy diffusion is it compatible or should I get automatic1111

Sometimes it activates, sometimes it doesn't.
Yes, it's in the right folder. Yes, I'm using the correct filename. I don't understand why, I don't modify the lora's name in the prompt, so if it works once, it should work all the time. I check the lora hashes in png info and some photos created in the same session have the lora hash, some don't, it's driving me crazy.
Does it have weird interactions with other loras? I'm using Adetailer, a face lora, a skin detail lora, dynamic thresholding and dynamic prompts/wildcards. I don't see why these would impact noise offset's activation.

I don't know how to use it I downloaded and use this lora but the software shows it is red and flashing I know he is not functioning does he need to use a trigger word to trigger so what is the trigger word

One of my favorite LoRAs! Are there any plans to port it to Flux?