Disclaimer: Heavily subject-tuned merge made from even more heavily subject-tuned models, and specializes in artwork depicting fantasy or sci-fi creature designs, especially anthropomorphic ones. While you may be able to diverge from this frame of reference, your results will likely not be as good as those to which the merge is attuned. Please make sure you download the .yaml config file and place it in your models folder, as this is required for the model do its thing.
The future of my previous projects was looking bleak, as each new version I tried to work on became more and more washed out. But no longer! I present the solution, and it's different enough that I'm marking it as a brand new merge series. Gone are those wild 28-30 CFG requirements, and in fact, this one's good at making images without artistic references, though it still greatly benefits from them.
My general recommendations remain the same; I highly recommend getting the ADetailer addon, as well as using Hires.fix to bring out better details, especially in characters' faces. Very versatile in my tests within its chief parameters (which are, bluntly, furry character designs), with less satisfying results when using more generic anime- or photorealism-style prompts, and I don't recommend this merge for generalized use. Can produce SFW and NSFW art, but may lean more heavily toward the latter without proper prompt guidance.
Description
A batch of general aesthetic updates. While a bit more heavily reliant on artistic style prompts, the results are generally superior, in my opinion.
FAQ
Comments (2)
What Image generator would you recommend for this model, preferably free ones, but paying is not a problem thanks in advance
tl;dr I don't have experience with any non-local generators for SD, so I don't personally have anything to recommend along those lines, whether free or premium services. While local installations of something like ForgeUI or A1111 are technically free, you generally need some reasonably robust hardware (graphics card in particular) to use them effectively, though it frees you from daily or monthly generation limits and imposed censorship most online generators seem to be so fond of. I recommend Forge, of the local UIs, but it needs a little effort to get "v-prediction" models, such as this one, to work, so read on below if you're interested in the solution.
I personally use a local installation (on my own PC) of Stable Diffusion, running the Forge UI; however, it's had some updates that make it no longer able to load the config files for models like these, and requires a specific 'branch' (or older version) of the UI to load them correctly. Having spent several frustrating, mind-bending, self-gaslighting weeks trying to fix a problem it seemed like nobody else on Earth was experiencing, I found and documented the fix over on Reddit, and if you go down the ForgeUI route, it should be helpful for you here as well, so I'm just gonna copy-pasta myself here.
If you're using Stable Diffusion with the Forge UI, it turns out that this month's (August 2024) batch of version commits to Forge have somehow, or for some reason (intentional? Unintentional?) made it unable to read the .yaml Config files required by "vpred" models to properly produce images.
This particular solution works with Forge:
Go to the root directory of your Forge installation (this includes if you have the very latest version capable of running Flux), which will be the folder where environment.bat, run.bat, and update.bat are located, as well as the folders "system" and "webui". In this folder, create a text document and copy-paste the following text:
@/echo off
set PATH=%DIR%\git\bin;%PATH%
git -C "%~dp0webui" checkout e3522c89191a01e0dd5855abbfd15cb685be3634
pause
(Stop at "pause") Now, using Save As, switch the file type in the dropdown box to All Files, and save it as something like rollback.bat. The string of text after "checkout" is a commit number, which you can find scrolling through the pages here https://github.com/lllyasviel/stable-diffusion-webui-forge/commits/main/ and navigating to a date that you're certain it was working. Run the new .bat file you created, and the CMD window will indicate that it is "switching" to a branch from the regular commit. If you need or want to return to the latest version of Forge, create a copy of your rollback.bat file and call it something like return.bat, and change the line after "checkout" to simply say "main" (without quotes.) Just run this .bat (double click on it) to return to the latest main version.
This should allow you to rollback the UI to a previous version, in this case, one where it was still able to read model config files - a must-have for me, since almost all of my model merges are apparently "vpred" models (I still don't understand what this is, or what it means, only that during creation, they spit out a .yaml file, and if that's not in the checkpoint folder with the model, it produces truly horrendous images - but the results when it works right are excellent, so I never complained.)
I do recommend that, if possible, and you have enough disk space, that you make a separate installation of SD Forge to run the older version with, along with any specific models that only work with that version, so that you don't have to keep switching the versions back and forth on one install directory (which I fear is likely to just mess things up somehow, eventually.)
There are also ways that, if you have multiple UIs (ComfyUI, Forge, A1111, Invoke, etc.) to set them all up to load checkpoint/model files, VAEs and such from the same directory, instead of having to take up hundreds of gigaybytes with multiple copies of models for every UI. Forge is very easy to install and set up by default.
Visit https://github.com/lllyasviel/stable-diffusion-webui-forge/tree/b58b0bd4259cf71077dfd7787fb77af4c02760a1 and scroll down to the heading "Installing Forge", then the link that says >>>Click here to download one-click package<<< and follow the instructions. The info in the block above would be how you roll back to a version able to load and use this particular model and other 'vpred' models using .yaml config files.
Details
Files
steinmix_charlie2.yaml
Mirrors
furationBETA_rebootedVPREDPartF.yaml
furationBETA_rebootedVPDModelbaseF.yaml
9thTail_mainV02.yaml
indigoFurryMix_se01Vpred.yaml
SD15ColorsplashV_v01.yaml
SD15ColorsplashV_v013.yaml
yiffyfluffybutts_V061.yaml
subfurryanalog_v10.yaml
easyfluff_5000DefaultvaeFFR3CS2.yaml
pegasusmix_VPredV1.yaml
9thTail_mainV01.yaml
9thTail_mainV03.yaml
beanv6flufusv2_v10.yaml
9thTail_altV03.yaml
steinmix_charlie2.yaml
steinmix_alpha6.yaml
yifurr_cid4.yaml
steinmix_alpha5.yaml
steinmix_alpha3.yaml
yiffymix_v41.yaml
steinmix_alpha6bL.yaml
indigoFurryMix_se02Vpred.yaml
yifurr_cid.yaml
furationBETA_rebootedREFINEDVPREDF.yaml
yifurr_sunrise.yaml
fluffyrockanimedilut_v10.yaml
fluffyrockE6LAION_e98E6laionE47.yaml
fluffyrockUnleashed_baseV10.yaml
yiffymix_v43.yaml
yiffymix_v44.yaml
pegasusmix_VPredV2.yaml
SD15ColorsplashV_v011.yaml
subfurryanalog_v22.yaml
subfurryanalog_v20.yaml
fluffyrock_e184VpredE157.yaml
fluffyrock_e257VpredE11.yaml
yiffymix_v40.yaml
yifurr_2.yaml
yifurr_sun.yaml
yifurr_newmoone2.yaml
yifurr_c3.yaml
9thTail_softV01.yaml
9thTail_juicyV03.yaml
SD15ColorsplashV_v012.yaml
fluffyrock_e233VpredE206.yaml
yiffymix_v42.yaml
9thTail_softV03.yaml
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.









