Go on and paw into the furry side of AI with NovelAI’s long-awaited NovelAI Diffusion Furry V3 image model now! We’re excited to see what amazing new things you will create with our image generation upgrade!
#NovelAI
Image Generation Progress Showcase:
When you include tags in your prompts, you may produce more of the same character with greater consistency!
NovelAI's Image Generation,
#NovelAIDiffusion
is live on 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.
The AI powered anime image generator to unleash your creativity: No boundaries - except the limits of your imagination.
#NovelAIDiffusionV3
is out now on , try it out!
あなたの創造力を解き放つ、AI搭載のアニメ画像ジェネレーター:
After a long day of generating, it's always nice to kick back under the warm sun and take a nap. Probably not a good idea to rest for long though, as our anime image generator AI
#NAIDiffusionV3
will be 𝗿𝗲𝗹𝗲𝗮𝘀𝗶𝗻𝗴 𝗲𝗮𝗿𝗹𝘆 𝗻𝗲𝘅𝘁 𝘄𝗲𝗲𝗸. Stay tuned!
@NilaierMusic
Since we are training on Danbooru, it also learns character names and their visuals. You can prompt for "masterpiece portrait of smiling rem, re zero, caustics, textile shading, high resolution illustration" and get this:
Whether it's the classic frames or trendy styles, there's something undeniably adorable about girls in glasses. Will you generate your perfect waifu in glasses with
#NAIDiffusionV3
releasing early this week?
[Announcement: Proprietary Software & Source Code Leaks]
Greetings, NovelAI Community.
On 10/6/2022, we experienced an unauthorized breach in the company's GitHub and secondary repositories.
The leak contained proprietary software and source code for the services we provide.
While working on the
#NovelAI
Image Generation, the team is creating images back and forth with their characters.
Tonight they are fighting with magic.
We won't have story-based
#NovelAI
Image Generation at launch, but two separate browser tabs (one for each character) + a third for your story should let you illustrate your tales quite easily!
The protagonist is trapped arguing in the swamps with his sassy traveling companion.
We still have to whip Pipkin Pippa knowledge into our text models BUT the image gen model knows Pippa!
The magic is in the tags: pipkin pippa, pipkin pippa (1st costume), phase connect, pipi hairclip
What a fun (and maybe a little yandere) stream so far already:
[Announcement: Proprietary Software & Source Code Leaks]
Greetings, NovelAI Community.
On 10/6/2022, we experienced an unauthorized breach in the company's GitHub and secondary repositories.
The leak contained proprietary software and source code for the services we provide.
[Update] NovelAI Diffusion Anime V3
We are happy to introduce you to our newest model:
#NAIDiffusionV3
Better knowledge, better consistency, better spatial understanding and it is even quite adept at drawing hands (finally!)
Full Release Post:
We're celebrating the collaboration with our favorite rabbit
@pipkinpippa
so it's coupon time! Use code PIPPA10 for a 10% discount at checkout for all Anlas purchases and NovelAI subscriptions from now until March 19th UTC!
Blog Post: NovelAI Improvements on Stable Diffusion
A very technical new blog post detailing some of our architecture and training developments made during the creation of our
#NovelAIDiffusion
Image Generation.
Let's talk about the magic behind
#NovelAIDiffusion
We’ve been quite busy watching NovelAI spread across social media, and we’re honestly surprised by the interesting theories people have come up with as to how our AI Image Generation tool works.
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.
[Update] Today, we're excited to announce a new set of
#NovelAIDiffusion
Image Generation Samplers: nai_smea & nai_smea_dyn - The Hi-res connoisseurs' choice as well as Higher Resolution Limits!
[Image Generation Update] Introducing
#NovelAI
Diffusion Anime V2
We are very happy to introduce you to the first new image generation model
#NAIDiffusionV2
. This one is mainly intended as an update to our current model and is still based on the same technology.
Since we've mentioned it in the
#NAIDiffusion
Image Generation Announcement, we've just sent all existing and active Tablet & Scroll subscriptions an additional 500 Anlas!
As mentioned previously, our new Image Generation Model further improves what we've learned from our Anime Module.
Who else needs an AI Generated Madoka Magica x Evangelion collab in their life?
#NovelAIDiffusion
Furry (Beta V1.2) has arrived!
We're excited to announce the second iteration of the Furry model, which generates even better content for furry and non-humans: monster girls, eldritch abominations, (cen)taurs, living vehicles, robots, elementals, etc!
Go wild!
In a world of unlimited possibilities which empress will you choose? You can generate them in different styles early next week with the
#NAIDiffusionV3
release.
無限の可能性が広がる世界で、どの女帝を選びますか?来週早々に、
#NAIDiffusionV3
Anlas Compensation Notice:
Greetings NovelAI Community!
We just deposited 1000 Anlas to all accounts that were subscribed after October 13th or that generated images between October 13th and October 18th.
General Update & Notes Regarding Unauthorized Breach, Generation Issues
Over the past week, you may have noticed delays in text generation, maintenance on Custom AI Modules, and general instability issues. These events are related to the recent internal security breach.
[New Image Generation Feature] Vibe Transfer
Today's release is a new feature for
#NovelAI
Diffusion V3 called Vibe Transfer. With Vibe Transfer, you can upload an input image that will be used as a sort of inspiration for generations while enabled. You can control the strength
Our third iteration of the
#NovelAIFurry
(Beta V1.3) Model is live!
Featuring further improvements to coherency, details and definition. Some generations were often monochromatic, so we fixed that!
Don't overlook the possibilities of the
#Furry
model - it can do so much more!
Updates on
#NovelAI
's latest image generation development: Our own model of
#stablediffusion
now knows how to resist some of the occasional strange cropping behaviors.
This helps the AI zero in on the intent of your prompt, regardless of the aspect ratios you're working with.
Goodbye Calliope
A while ago, we retired our oldest model, Calliope, which was based on EleutherAI’s GPT-Neo 2.7B. This model was our first step into finetuning large language models. We made some messes and learned a lot from the experience of creating her.
Right when
#NovelAI
It's time for another round of small
#NovelAI
updates!
Image Generation:
We've added the ability to reorder layers in the image editor! Now you can easily edit on different layers by reordering them.
Layer ordering makes edits easier than ever!
[Update] Inpaint with
#NovelAIDiffusion
!
Inpainting is a new tool that allows you to restore or regenerate specific areas in generations or uploaded images. It allows you to select a region of the image, and change what was already there.
@DougDougFood
"the only ones to accurately calculate how two video game characters would actually fight"... That's a lot of pressure! Thank you for putting your trust in NovelAI to carry out a part of such a critical task!
Great stream and video, especially those illustrations, wow!
Fun
#NovelAIDiffusion
tip: Just do a prompt as you would normally, then add {{{pixel art}}} to the tags!
You can also get better results by lowering the Scale Setting a little!
Tagging also makes it simple to alter the appearance of a character while keeping the same overall design idea in mind. Switch things up on the fly!
Male, female, well-endowed, less-endowed, thin, chubby, older, younger, hair ... no hair? It's all up to your preference!
#NovelAI
Image Generation Progress Showcase:
When you include tags in your prompts, you may produce more of the same character with greater consistency!
Our first Model
#NAILM3B
Clio is Coming to Opus!
Trained from scratch on our own custom datasets totaling over 1.5 trillion tokens (~6TB of text), is live for Opus subscribers right now.
Only 3 billion parameters, performing well above it’s level.
Introducing NovelAI-LM-13B-402k, Kayra
Technical Details:
NovelAI-LM-13B-402k Pretraining
If you are looking for the technical details, you can get your fill at our wandb report:
With our usual evaluation metrics, NovelAI-LM-13B-402k, our foundation
First off, the AI doesn’t have a database of images.
It has no image files to pull from or piece together.
Our AI is not a human-written algorithm that copies and mixes already existing images.
It does not photo bash.
It does not collage or match up existing visual data.
More neat furry themed
#NovelAI
wip image generations from Aero, our AI developer!
Using our own modified
#stablediffusion
model implementation merged with some of our already existing
#AIstorytellling
text generation tech.
You're in for a big Editor update today!
We completely overhauled the design of the action bar and Editor V2:
- Added a new generation animation on the updated Send button.
- Redesigned HypeBot and Tips to better match the new action bar.
Celebrating Three Years of imagined worlds!
We invite you to celebrate our third anniversary with us and learn about what we have planned for the future!
[NovelAI Tutorial]
#NAIDiffusion
Image Generation:
Creating Consistent
Characters [Female]
Learn how to use tags to create a female AI generated character who retains a consistent appearance across different images!
Official Blog Tutorial Post: