Tensor.Art
Create

NovelAI, NAI Diffusion V2

CHECKPOINT
Reprint

Updated:

Anime Safetensors
Anime

NovelAI Diffusion V2 Weights Release

Public NovelAI Diffusion V2 image generation model release As the field of image generation progresses, NovelAI has decided to release the weights of our second generation image generation model to the public, for the purposes of research, personal use and historical preservation. This way, users will be able to use the model even after it gets retired from our website. You can find the weights here:

While we are taking a trip down the memory lane, here are some of our relevant blog posts from back in the day, when we had just released this model on our website:

As written in the second of those blog articles, our model was trained using the output of the penultimate layer of the CLIP text encoder, so when using this model, make sure to set your image generation UI to use the second to last CLIP layer rather than the last one. V2 was also the first model we trained that was capable of generating resolutions of up to 1024 * 1024 pixels without the use of SMEA, so if you inference the model, make sure to run it at resolutions like 832x1216, 1024x1024 or 1216x832.

The model is released under a dual license, with both terms of the original CreativeML Open RAIL-M license and the Creative Commons BY-NC-SA 4.0 license applying to them. This allows non-commercial: use, redistribution and the creation of derivative works, as long as they are attributed properly and distributed under the same terms.

Of course, this model is a bit dated, so if you just want to have fun with a state of the art anime and furry image generation model, make sure to check out NovelAI V4.5 on our website and stay tuned for future updates! You can start for free with our new trial!

To reiterate, the weight release is only applicable to the second iteration of our NovelAI Diffusion model, all newer models remain proprietary and are only available via our image generation website.

NovelAI Diffusion V2 will be removed from novelai.net at a later date.

Safetensors version in https://huggingface.co/p1atdev/nai-anime-v2-safetensors

Version Detail

SD 1.5
This is our old, more capable SD1.5 based anime model. We are now making the weights publicly available for research and personal use. To keep the release compact, it only contains the Unet and VAE components. If you require the other components, you can copy them into the folder from the original Stable Diffusion 1.4 release. For convenience, are also providing a checkpoint in the classic ckpt format, but we have not verified its correctness. The model was trained using the text embeddings produced by CLIP's penultimate layer, so, depending on your inference software, you will need to set "CLIP skip" to 2. It supports 1 megapixel resolutions (e.g. 832x1216, 1024x1024 and 1216x832).

Project Permissions

Model reprinted from : https://huggingface.co/NovelAI/nai-anime-v2

Reprinted models are for communication and learning purposes only, not for commercial use. Original authors can contact us to transfer the models through our Discord channel --- #claim-models.

Related Posts