Anima version is available!
Nova Furry XL
A 2d/2.5d furry checkpoint model that can have great details on any type of furs, scales and feathers, which aims to be XL-fied Peaki Furry
Rules
You cannot use the generated images for commercial use if it's not edited (or just turning it to black and white)
Advertising is always welcome
Recommended settings:
Sampler: Euler a for best performance
Steps: 20-50 steps
(Pony)
CFG Scale: 6-7
Prompt: score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, source_furry, BREAK, furry, anthro, detailed face eyes and fur, {Animal Type}
Negative Prompt: blurry, jpeg artifacts, username, watermark, signature, normal quality, worst quality, large head, low quality, text, error, missing fingers, extra digits, fewer digits, bad eye
(Illustrious)
CFG Scale: 3-5
Prompt: masterpiece, best quality, amazing quality, very aesthetic, high resolution, ultra-detailed, absurdres, newest, scenery, furry, anthro, {Prompt}, BREAK, depth of field, detailed fluffy fur, volumetric lighting
Negative Prompt: human, multiple tails, modern, recent, old, oldest, graphic, cartoon, text, painting, crayon, graphite, abstract, glitch, deformed, mutated, ugly, disfigured, long body, lowres, bad anatomy, bad hands, missing fingers, extra digits, fewer digits, cropped, very displeasing, (worst quality, bad quality:1.2), bad anatomy, sketch, jpeg artifacts, signature, watermark, username, simple background, conjoined, bad ai-generated
Description
Potential fix about details on further distance (eg. eyes)
Better posing, fur details and structures
FAQ
Comments (45)
She's beautiful. But for me personally there's one small downside, I realize it's a model for furries, but it's a shame she can't do humans in the same framework as a furry. Say, if I want to depict a sexual act between a human and a furry, I'll never be able to do it.
What's new with 2.0?
Poses, eye details and as someone said on v1.0, the prompt leaking is much lesser
A friend recommended this, and my first try did not disappoint! Thanks for making such an awesome furry checkpoint! I will leave a review soon, once I have had more time to try things out.
This is by far the best model I've ever used. I thank my friend for introducing me to this. You've done some great work here. Keep it up!
Sorry for this possible dumb question but I use NovaAnime to generate furry art as it has lots of character tags from danbooru and e621 which is awesome,
does NovaFurry also have this and does it support more anthro species/characters than NovaAnime?
Yes, NovaFurry also have plenty of tags and also supports more species/characters for furry and much better fur detail than NovaAnime in my opinion
@Crody Does NovaFurry have a full tag list for people to share? Cus people already made wildcard e621 megalists with 500K+ tags
@TomLucidor Unfortunately no but I guess you can just use e621 for that
@Crody How many images did you make this with? Trying to sense how diverse it can be
@TomLucidor I made around 50 images for the showcase, more than 200 images for the experiment and double numbered images for personal uses
From the experiences, I can guarantee that most character tags including species works well
Other tags such as pose, expression, position and interaction works well too but I cannot guarantee for the tags that doesn't have enough images uploaded on e621
@Crody I meant the data source for the Checkpoint, cus larger diversity~ better?
@TomLucidor I don't know fully about the source because I merged LoRAs to cover them
@Crody This would become a hard game of "sauce tracing" for the major Checkpoints used in this (rather than "just the LoRA")
@TomLucidor I don't know this would be the proper answer but I don't want anyone to repeat the same kind of thing happened on ChilloutMix
@Crody Wait what happened there?
@TomLucidor History time: Around 2 years ago, free use ChilloutMix becomes property of others because of take down of the original
With the second uploading, the rule changed to have restriction on merging and image sharing thus some models like AbyssOrangeMix3A1 which includes it in merge became prohibited to share / monetize images across internet
To avoid that, some user made almost identical model or straight up same model but labeling as different model and share it as free use model
What I don't want is that making same model over and over
I just want everyone to make their own model, not copy
@Crody Yeah that is grim cus I would rather aggregate the best unique models, than seeing the same models being overused. Also not enough sharing of scaled/auto dataset making, along with the bar of entry in GPU power.
Fantastic model and one of the absolute best go-to models for Furry generation. I am super happy that it also seems to work reasonably well with pony-based SDXL loRAs
Illustrious v2 is pretty versatile and the output is really good, but there are many problems with shot sizes, and most of the time the camera angle is really weird
Could you make this available for mobile?
You can use it on PixAI
Browser version is the best
Astonishing, outstanding model, literally the best choice among furry models for illustrious. In all the time i've been using it, i have yet to encounter an unrealized concept problem.
Hey, would you look at that - it's a model that does even more deranged shit than the usual PonyXL6, and has slightly less bizarre prompting.
Let's get this out onto a tray ..
Nice!
nice hiss
Illustrious v2 handles almost as easy as SD1.5, yet has the knowledge and vocabulary of SDXL. I feel like it knows every character I want without any character LoRAs. Probably the best model I've tried so far. The only bad thing is that it has a hardwired style that I wasn't able to modify through my prompts.
A consistent visual style is more in the realm of funny IPAdapter setups (or some other, future toy) finalizing the output of a separate compositional model (such as this one), for anyone passing by this comment.
This model is FAR better than most at handling a low step/CFG setting for quick draft results that you can then img2img improve on normal settings. I generally use 8 steps and CFG of 2 in order to generate my first pass results, and they're nearly as good as the final process.
It's true, so you made me want to make a minimalistic ComfyUI setup to do this, while abusing "SPO-sdxl" to make the resulting image sparkly.
here you go;
https://civitai.com/images/49381723
(workflow attached to PNG, or in the clickable Nodes button in the "Other metadata" section of the civitai image view.)
Semantic-aware Guidance optional while prompt-surfing, though definitely beneficial when fishing for "final compositions".
so much of your recommended prompt is full of token bloat. there's no meaningful difference between your "recommended prompt" and just using masterpiece, best quality, and then your prompt.
The recoomended prompts are really weird, because in my experience - the model works very well using super simple descriptions and barely any negatives.
@VeerGeer less is more
@chuunikaede Well, doubly so these days - when you can simply use something like SPO* for taking an output image, and re-cooking it with a more sparkly look.
* https://civitai.com/models/510261?modelVersionId=567119
For fixing fuzzy lines and low facial detail, one can also "just" cook the image with
Flux-Fusion-DS-merge
which provides surprisingly good results after 2-4 steps at denoise values that depend on the implementation ('simple' scheduler, 0.22 for invokeai at 3 steps, 0.3 for comfyui)
scaling the image up, performing this weird flux refinement, and scaling it back down also has its perks
due to the fact that you're cooking an existing image, the input prompt doesn't really matter - so you don't need a high quality text encoder - so the nf4 or Q4_0 gguf versions do the job just fine.
@VeerGeer i have no idea what any of this means lol.
@VeerGeer hey, actually amazing info, thx mate
Can I ask you question for all your model?
Are monetizations allowed to support accounts like Buzz or tensorART's Buffet plans?
(It's not about monetizing the model itself.)
Also, I found artist names are not working properly. I added following artist name at the beginning of my prompt but It's now working. can someone tell me how to use artist's artstyle correctly?
, jinxit, oro96, omurice, woolrool, rabbitadvisory, theobrobine
EDIT: Artist styles work fine, you need to put them in parenthesis as such.
masterpiece, (by woolrool), anthro and
(by woolrool:1.4) for stronger effect.
Keep note that some artist names change, and some weren't ever trained in the model because they're way too new (rabbitadvisory for example.). Test at your own leisure.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.