Thread by @novelaiofficial

NovelAI's Image Generation, #NovelAIDiffusion is live on novelai.net now!

NovelAI Diffusion Anime image generation is uniquely tailored to give you a creative tool to visualize your visions without limitations, allowing you to paint the stories of your imagination.

We decided to release #NovelAIDiffusion models as standalone releases:

NovelAI Diffusion Anime (Curated)
Good baseline quality and predictable subject matter.

NovelAI Diffusion Anime (Full)
The expanded training set allows for a wider variety of generations.

In addition to the #NovelAIDiffusion Anime models, we are also shipping with the NovelAI Diffusion Furry (Beta) Model:
Easily create Furry and Anthropomorphic Animal themed content.
This model is not a complete work and will be expanded upon as development proceeds.

Let's focus on what makes NovelAI Diffusion so unique!

Tagging:
The team seized the opportunity to fill a gap in existing Image Models knowledge and teach the AI about tags!

When you type your image generation prompt, tagging suggestions will appear.

We’ve increased the CLIP token context capabilities from 77 tokens to 231 tokens, giving you more space to craft your prompt than ever.

The Model has been trained without crops, which means now you can generate arbitrary aspect ratio images. Lastly we’ve greatly improved the language understanding of the CLIP model.

Our finetuned #NovelAIDiffusion allows you to give the AI clear instructions on what to generate.
The AI will suggest tags based on what you type and display corresponding circle markers indicating how much knowledge it has of each tag.

The use of tags is not required, but they can play a critical role in steering the AI in the direction you want. We highly recommend a mix of tags and normal text prose. If used to create a 'character' of your own, you can create the illusion of consistency across generations.

Pricing & Features:
For Opus Subscribers: We will offer access to unlimited basic generations at up to 640 x 640 pixels at no additional cost.

This includes all Image Resolutions under the “Normal” and “Small” dropdown categories.

The Opus tier offers unlimited generations with select settings, while the lower tiers are perfect for experimenting to find out if NovelAI's Image Generation service fits your needs.

Pricing is highly dependent on resolution size, the number of diffusion steps used, and the number of images generated in a single generation.
Features and advancements are still being made, and we will be listening to your feedback as we continue to develop.

The New Currency
We've taken this opportunity to rebrand our currency, inspired by our company name, Anlatan, with the future in mind.

One currency, used everywhere, introducing: Anlas.

All subscription tiers will be able to use Anlas to access our image generation and its features.

We have also raised the default monthly amount of Anlas to 1000 for both Scroll and Tablet Subscriptions, allowing use for Image Generation in addition to Custom AI Module training.

Please note that an active subscription is required to purchase Anlas, which are needed to generate Images with #NovelAIDiffusion.

With that out, let's get on to some features of our Image Generation!

All subscribers will have access to image-to-image (img2img) features, in which you can tweak, modify, or even draw images from scratch and have the model generate from it as a base.

There are many others: Edit Image, Enhance Image, Generate Variations, Undesired Content and Strengthening & Weakening Vectors all accompany our tag-wielding models!

You can read all about the features over on our blog and our documentation page: docs.novelai.net/image.html

So, what happened to the delays?
We've heard your feedback on our delay announcement loud and clear.

Our users want an experience with unprecedented levels of unfiltered freedom, and so do we.

We have come to the decision to only offer our own unique models and forego the basic Standard Diffusion model, which is already offered in many other places and also would have required us to pay close attention to what kinds of content its different versions can produce.

By offering our own exclusive experiences, we're able to provide you with a tool designed for visual storytelling without any limitations.

NovelAI is meant to be a storyteller, and we believe our Image Generation will allow you to paint all kinds of stories. We are only beginning to expand the capabilities of this feature. So, at this time, we no longer plan to provide the basic, standard Stable Diffusion model.

Make sure to test and see if our NovelAI Diffusion models suit your needs.

Please keep in mind that the NovelAI Diffusion models will not produce the same results as the standard Stable Diffusion model, even with the same prompt and seed.

Copyright & Usage Rights
Images uploaded and generated on NovelAI are not stored on our servers. Make sure to regularly backup and save your images and text prompts.
We do not act as a platform to share images. Please use other services in order to share your generated images.

NovelAI does not claim copyright protections or ownership of any files. Images are treated as Content as defined in Section 1.4 in the Terms of Service. You must conduct your own research if you want to use your generated images for any settings outside NovelAI.

Text prompts, image generations, canvas edits, and image uploads are not stored and will be lost at the end of each session.

Please generate responsibly, and be mindful of your wordage. Image Generation is a new technology that may generate unpredictable results.

Feel free to tag us, and your generations with #NovelAI or #NAIDiffusion so we can follow the journey of your creations!

If you have any issues, don’t hesitate to contact [e-mail]

The team wishes you tons of fun with #NovelAIDiffusion!