In search of something, we sometimes lose our way. Where is this going? (I don't mean this in a bad way; I like your models)

>Enhancing variety of species

New species? What are they?

I've been doing a whole lot of testing with previous models you've made, and have found that over the iterations a lot of edge-case tags stop working entirely or get strongly diluted down. For example, artist tags that got responses in earlier model stop getting responses at all in the newer versions.

Is there any particular reason why this might be happening?

Hi, I'm just stopping by to ask you a few things because I'm about to open a page on X and Patreon, and I'm going to use your checkpoint merge to generate the images, I'm going to edit them with style loras and manual editing using Picsart, and I'm thinking about making subscriptions on Patreon, I'd like to know what the clauses are so I can create paid subscriptions, could you tell me so I don't infringe your copyright? I already have a banner made, would you like me to send it to you?

Armadillo girl was impressive I had no idea what to expect.

I did a test of 69 species of furry. Furry like mammals only in my list https://civitai.com/posts/19279634 https://civitai.com/posts/19279747 https://civitai.com/posts/19279779 https://civitai.com/posts/19279799

I had more in the list but "female goatman" did nothing for example.

Most of them did great I think. But it wasn't super fun doing it, someone else can check the birds, fish, aquatic mammals, insects, reptiles and more fantasy creatures.

If anyone can figure out what I need to say to make them sit in the damn innertubes I'd be grateful. Also to make the innertube at least be touching the water!

si

WE ARE SO BACK (That implies we lost but whatever) YOU SIRE ARE THE GOAT

what is the difference between b and a versions???

After having done a whole lot of tests with 9B, I've found the model to be a lot less flexible than 7B.

I think 9B would be more useful for people who don't like to try weird ideas and just want the straight-away "small simple prompt with popular species get pretty picture" approach.

But 7B currently remains the strongest model in this series that has better capacity for handling less "common" concepts/furry species.

I watched too many Disney and DonBluth movies growing up so cute mice and cute 'chipmonkey' gals appeal to me. It's too late for me. Save Yourselves...

Thank You. Superb Work.

hola!

Is there a list of artists prompts?

I ran some comparions between the last six versions. Results are here:

https://civitai.com/articles/16831/nova-furry-style-prompt-adherence-testing

Just I can say this is one of the best models,you can make good stuff and simple

is there a list of artists prompts?

the best

what about A and B? 🤔

Heavily trained on females. Even when prompting for males, they tend to end up with female bodies and breasts.

Poor eye generation from the distance (as little as a full-body character on a 1024x1356 sheet), anything that can be done with it?

I have a problem. When generating with LoRa, the images are completely blurry.

Not sure how to explain, but last versions looks more like... AI? While first versions had some style and uniqueness in them

But it is still the best furry checkpoint, no doubt

Anybody have a workaround or some tips on how to use the FaceDetailer node on comfyui for Furry/Anthro's? I might need to swap the segment model or something, since my outputs are getting their faces fixed (obviously) but turning them human like, I do like the output but wanted to know if there was a specific setting for furries.

Bring back the older versions for the in-site gen gang

Latests versions are good by themselves, but they lack of that flexibility that the older ones had. The style itself messes up pretty much every style you put as an extra (shading, over contrasted, AI-ish) It just lost that 2d cartoony side :(

Unfortunately I can't use this one.
See, I was using a few particular characters like Mew but this new version is downright incapable of getting it right.
Over time, it seems like this model got worse and worse, I think version 7b was peak.

"You cannot use the generated images for commercial use if it's not edited (or just turning it to black and white)"

bro who do you think you are lmao, YOU DIDNT EVEN CREATE THESE IMAGES you fat basement dweller Twat.

When comparing to my older renders e.g. with v6, I also notice that the newer style seems more dull and less vivid. It had more crisp and contrasty colors and outlines before, and it could be wonderfully merged with NovaAnimalXL to create nice hybrid 2.5D looks. Also, it seems that with each iteration, the models gets worse in creating realistic macrofurry scenes. Buildings mostly look square and boring, adding keywords like "destruction, damaged buildings, cracked asphalt under foot, burning buildings, buildings on fire" seem to work less well to get the desired effect. In older renders, when a macro was sitting on a building, the structure was crumbling in an almoist realistic manner, now I can only replicate it with img2img. Please train more macro data, since it's a very popular fetish, and it benefits good-looking cities and towns as a neat side effect. Thank you 🙏 If you drop me a PM, I can provide you visual examples. (I am back to V6b now and happy again)

Would really appreciate if we could use the older models in browser, some people don't have access to stuff like stable diffusion on mobile.

question, why is mine just giving a black screen

I'm having an issue with eye clarity and sometimes other parts of the picture being unclear, and I was just wondering whether it had something to do with the workflow I was using?

I've currently tried the SDXL Simple workflow in ComfyUI the Basic txt2img (w/ added clip skip), but I feel like there is something I'm seriously missing.