Sign In

Illuminutty Diffusion

898
6.6k
21
Updated: Oct 5, 2024
base modelnutty
Verified:
SafeTensor
Type
Checkpoint Trained
Stats
6,578
Reviews
Published
Apr 11, 2023
Base Model
SD 2.1 768
Hash
AutoV2
D1406B31C0
default creator card background decoration
OZ
ozreac

Update (April 12th, 2023): Apparently this model barely works without a set of embeds. Anyway, here's all of them. As with the checkpoint, it was licensed to me and I'm free to redistribute it. But even if you disagree: different name, different hash == different embeds.

Have fun.

https://files.catbox.moe/0gjduq.zip

--

You may notice a distinct similarity in the output this model generates with a model that was formerly available named 'Illuminati Diffusion'. That is because the weights and configs are identical. However, this is not Illuminati Diffusion v11. That name has been exclusively licensed to one of those shitty SaaS generation services.

In addition, although the weights and configs are identical, the hashes of the files are different. Therefore: different name, different hash, different model.

But still, let's make two things very clear:

  1. Weights and configuration files are not copyrightable, and;

  2. Even if they were, the weights and models were downloaded by me (and thousands of others) under the creativeml-openrail-m license.

That license grants me:

a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable copyright license to reproduce, prepare, publicly display, publicly perform, sublicense, and distribute the Complementary Material, the Model, and Derivatives of the Model.

And no, adding a blurb to the model page that no-one reads does not, in fact, constitute a modification of the license that you uploaded under, or amend the terms of service you were subject to when you uploaded it.

Given the above, I republished a wholly new work containing the original weights. You're now free to use this wildly overtrained model for any purpose you like. I suggest using it as an example of the need for ground truth datasets in training.