Use e621 tags (no underscore), Artist tag very effective in YiffyMix.
GridList Species/Artist(v64) update!! & LoRAs (SDXL)/samples/wildcards
Recommend Artist tags (NoobXL) & comfyUI workflow.
Example Setting (Use recommended sampler to get correct quality)
Setting SDXL (SDXL-lightning & NoobXL)
Steps = 12~24
Sampler = "DDPM", "Euler A SGMUniform", "Euler SGMUniform"
CFG scale = 3~4
Negative embeddings SDXL = ac-neg1, ac-neg2 (You don't really need this)
Postive LoRA SDXL = SeaArt Quality Tags LoRA (You don't really need this)
Stop at CLIP layers = 2
Setting SD 1.5 + vpred
Steps = 30~40
Sampler = "DPM++ 2M Karras", "DPM++ SDE Karras", "DDIM", "UniPC"
CFG scale = 6~8
Negative embeddings = deformity_v6, bwu [SD-WebUI\embeddings]
Stop at CLIP layers = 1
SD VAE = kl-f8-anime2, Furception [SD-WebUI\models\VAE]
Hires. fix
Hires steps = Steps * Denoising strength
Denoising strength = 0.25
Hires upscaler = 4x-UltraMix_Smooth [SD-WebUI\models\ESRGAN]
ControlNet
ControlNet = softedge_hed, control_v11p_sd15_softedge
ControlNet SDXL = softedge_hed, sdxlsoftedge-dexined, noobai-xl-controlnet
ControlNet Weight = 0.35~0.5
ControlNet Pixel Perfect = true
Wan 2.2 ComfyUI Animeted [workflow]
model = wan2.2_ti2v_5B_fp16
vae = wan2.2_vae
lora (turbo) = lora_wan2.2_ti2v_5B_turbo_lora_rank_64_fp16
lora (furry-nsfw) = wan2.2-5B_furry-nsfw-v2.0-e83
lora (livewallpaper) = wan2.2-5B_livewallpaper-720p
sampler = steps:4, cfg:1, sampler:"eular_ancestral", schedule:"sgm_uniform"
SD WebUI
LoRA Training SDXL
imgs count = 15~50
total steps = epoch * imgs count * folder loop = 3000~4500
network_dim = 64
network_alpha = 128 (SDXL) \ 16 (Noob)
learning_rate = 0.0002~0.0005
unet_lr = 0.0001 #learning_rate/2
text_encoder_lr = 0.00005 #learning_rate/4
lr_scheduler = "cosine_with_restarts"
mixed_precision = "bf16"
optimizer_type = "Adafactor"
optimizer_args = [ "scale_parameter=False", "relative_step=False", "warmup_init=False", ]
YiffyMix v4x V-pred Setting
# YiffyMix v4x A1111 SD-WebUI or Forge V-pred Setting
1.Download config file ".yaml" paste next to the model.
2.Raname ".yaml" same as model name. ( check the ".yaml" has [parameterization: "v"] line )
3.Restart SD-WebUI ( if config fails to load, the model just generate noise )
# YiffyMix v4x ComfyUI or StableSwarmUI V-pred Setting
Add node "ModelSamplingDiscrete", Sampling = "v_prediction"
# ComfyUI V-pred Setting (old version)
1.Copy config ".yaml" to [ComfyUI\models\configs] and refresh ComfyUI
2."Load Checkpoint (With Config)" [Right Click\AddNode\advanced\model_merging]
workflow:[Load Checkpoint (With Config)]-[KSampler]-[VAE Decode]-[Save Image]
# v-pred mode troubleshoot
If you use new version webui-forge and failed detect v-pred model.
Try update webui-forge (click update.bat).
When loading v-pred model, you will see this line in cmd console:
left over keys: dict_keys(['v_pred'])
# v4.x some time will get little fried issue image in this case
Use "Dynamic Prompts with __wildcards__ prompts" and "batch size > 1"
Just generate again with same prompts with parament, it will return normal.
Version Info
v1.x [2D,512~768] old model, unstable, low resolution
v2.x [2D,512~896] e621 dataset + 30~40% SD5 dataset
v3.0 [2D,3D,512~1088] large dataset then v2.x version, can do 2D, 3D
v3.1 [2D,3D,Real,512~1088]
more realistic, but loss some concept (e621 tag count lower 1000)
v3.2 [2D,3D,Real,512~1088]
unstable version, use SNR version FluffyRock, more noise detail
v3.3 [2D,3D,Real,512~1088]
stable version, more detail, more sensitive to prompts
v3.4 [2D,3D,Real,512~1088] ※include Fluffy Rock Quality Tags-LoRA
stable version, more detail, clear result, reduce some noise (like bush,pattern)
v3.5 [2D,3D,Real,512~1088]
unstable version, contrast & detailed enhance
v3.6 [2D,3D,Real,512~1088]
stable version, more sensitive to e621 tags
v3.7 [2D,3D,Real,512~1088]
stable version, reduce noise overactivity, fix blurry eyes and finger error, fix failed background depth
v4.0 [2D,3D,Real,512~1088]
v-pred version, remix from EasyFluff
accuracy anatomy and style with fewer prompt,
little dim blue average color with smooth noise, weakly negative prompt issue
v4.1 [2D,3D,Real,512~1088]
v-pred version, color and low contrast fix, reduced realistic noise.
v4.2 [2D,3D,Real,512~1088]
v-pred version, more contrast and detail, yelow issue.
if you feel too fried use Rescale CFG =0.35
v4.3 [2D,3D,Real,512~1088]
v-pred version, clear artist style, fix most yellow/brown issue
the version doesn't need CFG Rescale
v4.4 [2D,3D,Real,512~1088]
v-pred version, no more yellow/brown issue.
the version doesn't need CFG Rescale
v5.0 [2D,3D,Real,896~1536]
SDXL-lightning version, base on Compassmix XL.
v5.1 [2D,3D,Real,896~1536]
SDXL-lightning version, more e621 dataset, lower human face, better NSFW stuff.
v5.2 [2D,3D,Real,896~1536]
SDXL-lightning version, increase average quality. Little unstable than v51 but more creative.
v6.0 [2D,3D,res:896~1536]
More character better sex pose, limited and uncontrollable style (mostlly anime style).
v6.1 [2D,3D,Real,896~1536]
More realistic detail, reduce anime style, fix flat and boring background.
v6.1a-RE [2D,3D,Real,896~1536]
Same as v61 but adjust average style to simi-realistic.
v6.2 [2D,3D,Real,896~1536]
more effective prompt, keep style also with good realistic detail, saturation down little bit.
v6.3 [2D,3D,Real,896~1536]
Detail(noise) level between v62 and v61, upgrade character/artist accuracy.
v6.4 [2D,3D,Real,896~1536]
Upgrade light quality and average detail.
Note about NoobXL to create stable furry (v6x):
use "furry" tag in propmt can let AI create furry not human.
use "no human" tag in propmt can avoid NoobXL model add human and reduce anthro affect (more original style).
use "anime style" in negative propmt can reduce classic booru style and more realistic.
Some SD prompt trick:
Combine two character:
characterA \(characterB\)
Avoid tag bleeding:
(chain:0)-link fence
(cowboy:0) shot
high (collar:0)
Multi-tag combine, enhance and reduce token use:
from side + side view = from side view
crossed legs + legs up = crossed legs up
Basic Style
Negative Prompt SD1.5
unusual anatomy, mutilated, malformed, watermark,
amputee, mosaic censorship, sketch, monochrome
Negative Prompt SDXL
malformed, worst quality, bad quality, signature, text, url
3D Artwork Style
Prompt v5x: blender \(software\), ray tracing, 3d, unreal engine
Photorealistic Style SDXL
Prompt v5x:
by Mandy Disher, by Wim Wenders, by Robert Rauschenberg, [by Fossa666::0.65],
ultra realistic, photorealism, photograph
Negative prompt v5x: sketch, manga, vector, line art, toony
Prompt v6x: film photography, photorealistic, film grain
Negative prompt v6x: anime style, vibrant, pastel
Photorealistic Style SD1.5
Prompt sd15: [:photorealistic, analog style, realistic, photorealism:0.5]
Negative prompt sd15: [:bwu:0.5]
Recommend Refiner: IndigoFurryMix v110-Realistic Switch at 0.5
Description
Recipes v42:
partA = (FluffyRock-zsnr-vpred-e256-minsnr-e30 + YiffyMix v36 - FluffyRock-e99-snr-e72) * trainDiff:1.0
partB = partA + Polyfur-vpred-e66-snr-e15 * 0.45
partC = (partB + IndigoFurryMix v95 - SD5) trainDiff:0.35 + YiffyMix v40 * 0.5
partD = (partC + Dreamshaper v8 - SD5) trainDiff:0.2 + LoRA:[Detail Tweaker * 0.2]
YiffyMix v42 = partD + CLIP:[ReV Animated * 0.2 + YiffyMix v36 * 0.27 + EasyFluff v10-Funner * 0.1]
FAQ
Comments (47)
For when a Pony based XL version?
It would be interesting to create an SDXL/Pony version.
Maybe this is an issue specific to Invoke, but when I try to run v41 and v42 with the config in the same folder as the model I just get runtime errors. v36 works fine still at least.
RuntimeError: Input dtypes must be the same, got: input c10::Half, batch1: float, batch2: float [2024-03-22 16:36:57,758]::[InvokeAI]::ERROR --> Error while invoking: Input dtypes must be the same, got: input c10::Half, batch1: float, batch2: floatMost common settings for SD-A1111 or SD-Forge.
Only SD2.x and some SD1.x furry models use v-prediction.
But SD2.x vpred setting can't use in SD1.x, Sometimes problem from here.
There's a InvokeAI vpred branch (didn't test) .
You can try SD-Forge, same SD-A1111 interface and more faster less vram use.
(include python+git and bunch of builtin extensions)
Which of the VAEs should I use? There are two...
WD-KL-F8-Anime2 already baked in v4.x, but I recommend Furception
@chilon249 If it's already baked in, does it get overridden when I select another VAE in the settings or are they both used at the same time somehow?
VAE setting "None" or "Automatic" = use baked VAE
select another VAE will override it.
@chilon249 Ok, thanks.
Small correction: if you rename downloaded VAE as a model name and place it in folder with model, then, if you choose "Automatic" setting, will be used external VAE and not baked.
Any chance for a pony XL finetuning using your dataset?
YiffyMix is a merge, not a finetune. It relies on other models being trained in order to merge. There are few XL models and Pony is mostly incompatible with them, so I wouldn't expect a merge anytime soon.
# V-prediction model has following feature #
Low-Rescale CFG : low frequency noise, cluster detail, high contrast
Sometimes less detail but smooth surface color and less mutation.
High-Rescale CFG : high frequency noise, flat detail, low contrast
more detail can be shown, but not like EPS (non-vpred).
You can see the visible halftone dot in "starry night sky" or "grassland".
try "halftone background" to test it.
why does it looks so fried
you need the config file
(SOLVED by turning off "Disable memmapping for loading .safetensors files") v42 NOT WORKING(help me pls). ValueError: both buffer length (0) and count (-1) must not be 0 How to make it work? v41 works but v42 doesn’t, what else is needed for it besides the model itself and .yaml? I have WebUI v1.7.0 The error appears when I try to run the model, the launch is interrupted and the model does not change to YiffyMixv42
Are you using SD-WebUI a1111?
If you can't load new model, try remove "cache.json" and restart SD-WebUI.
SD-WebUI will rebuild model list.
@chilon249 V42 doesn't work for me either. I tried both A1111 and forge webui. I deleted cache.json, but it still doesn't work.
Turning off "Disable memmapping for loading .safetensors files" in the Webui did the trick for me to fix this bug
@beglueckendmaerc9974 thx! That works for me.
@beglueckendmaerc9974 Thank you so much!!!!! Just turning it off and everythig working fine now
ComfyUI, exact same error.
to fix this needs to revert patch described in https://github.com/comfyanonymous/ComfyUI/issues/1992#issuecomment-1914873427
I’ve used YiffyMix pretty extensively and have been pretty amazed with what it comes up with. It has the normal issues and take some patience and practice with getting settings just right and wording what you’re so so but so far I’ve not been disappointed.
Is there a practical way to generate multiple characters?
[2 characters]
duo, duo focus
[3 characters]
trio, trio focus
[2~n characters] (num control by width, best on front view)
lineup, comparing, group, group picture
[n characters] (need larger image)
large group
[1~n characters multiple position]
multiple positions, multiple poses, model sheet
[multiple frame] (need larger image)
2 panel comic, 3 panel comic, 4 panel comic, character collage,
three frame image, four frame image, five frame image, six frame image,
seven frame image, eight frame image, nine frame image, ten frame image
Can anyone tell me what prompt words to write so that the vagina is closed and not a huge hole?
Try putting "open pussy, gaping pussy" in the negative section.
@OspreyHawk Incredible,thank you!!Apparently I didn’t understand the principle of the prompts very well. After all, I wrote closed pussy in the prompt. Why do I need to make a negative prompt for an open pussy for it to work?
"barely visible pussy" should work too
Love the model
Please help me out with the yellowing issue, I have downloaded the config file and out nearby the model but it doesn't help at all.
What should I do ?
This yellow effect ruins all artworks.
This model is kickass, I love how I can mix different artist styles to get an eye-candy result. I use v3.6 because newer versions tend to have a weird yellow-ish tint which I don't like. The only significant downside is total incompatibility with AnimateDiff - spent a whole day trying to get it to work to no avail.
How can i fix that?
https://civitai.com/posts/2162743
Try removing "detailed pussy" from your prompt and moving your negative embeddings into the beginning of your negative prompt. For example start your neg. prompt with "boring_e621_fluffyrock_v4, EasyNegativeV2 verybadimagenegative_v1.3 ng_deepnegative_v1_75t" and then goes the rest. Also try adding "mutilated genitals" to your neg.prompt.
@oaf40 This didn’t help, what other options might there be? By the way, can the emebedding be mixed? Do they not interfere with each other? Can you reproduce my problem or does your vagina look normal in this case?
@oaf40 https://civitai.com/posts/2163625
Now it looks like this, thanks for the tips!
Can you suggest a word that would allow us to bring the “camera” closer to the pelvis? So that it would look like she are looking down and the camera is sort of between her legs. I am not a native English speaker
@VladossX I'm afraid I won't be able to reproduce your picture because I will face multiple issues, I am running SD on an old AMD videocard and I've explicitly set the "Random number generator source" setting to "CPU" while most people have it set to "GPU" which is not cross-platform and won't yield the same image as yours. And generating a 1024x1024 picture will take at least half an hour. Sorry for not being able to help this way. One of the options you can try is to download some pussy-specific LoRA (search for "pussy" on civitai) and enable it in your prompt.
@VladossX
> Can you suggest a word that would allow us to bring the “camera” closer to the pelvis?
try the "foreshortening" token with increased weight and specify "pussy in focus" as well.
@oaf40
>One of the options you can try is to download some pussy-specific LoRA (search for "pussy" on civitai) and enable it in your prompt.
I thought about it, but it's mostly human lora or anime, is this suitable for furries?
@VladossX yes usually it should work, try playing around with the weight of the LoRA. If it looks too "human", decrease the weight. Also I've just remembered,
> Can you suggest a word that would allow us to bring the “camera” closer to the pelvis? So that it would look like she are looking down and the camera is sort of between her legs
Besides "foreshortening" and "pussy in focus", try adding "low angle" or "worm's-eye view". To make your character look down at the "camera" use "looking down, looking at viewer". I recommend you to read this page on angles and perspectives https://e621.net/wiki_pages/10866 , all tags from this page should work in yiffymix model.
@oaf40
Sorry to bother you, but what word can remove hands? It produces a picture like this, https://civitai.com/images/9953995 I don't need hands reaching for the "camera"
Thanks in advance for your help. I won't bother you anymore!
@VladossX You can make her use her hands in some other way instead of reaching the viewer. You can check out my old post https://civitai.com/posts/1955266 and see the tokens I've used for that: "(breasts, fondling, holding breasts)". You might also want to try adding "(reaching towards viewer)" to the neg.prompt.
> I won't bother you anymore!
There's no problem with that, I will try to help if I can ;) Cheers!
Excellent model, but I have absolutely no luck with getting v_42 rid of the yellow tint. No matter what I try.
I do have:
- SD - Forge
- loaded ".yaml" config (at least console says "left over keys: dict_keys(['v_pred'])" when loading model)
I have tried:
- Rescale CFG plugin: (0.0-1 with not much success, Auto Color Fix did improve things a tiny bit, but it is inconsistent, and when doing batch size, it applies barely in the first image, much stronger on the second etc..)
- LatentModifier Integrated (Rescale Cfg Phi) or whatever is there with no desired effect, the higher number of Phi introduces unwanted fractures and artefacts (colourwise it is ok thought)
- whatever embedding, VAE, or low-contrast prompt there is.
Anybody has found a solution? Or have a tip? Thanks.
Yes, v42 has yellow color issue, will be fixed in new version.
@chilon249 Much appreciated, thanks for your hard work ⸜(*ˊᗜˋ*)⸝
@chilon249 Am I correct to understand that in v3.6 only the artists listed in the "Species/Artists grid list" are recognized by the model? For example I'd like to use rz54's and skorpdk's style but they aren't in the list, does the model know their styles nevertheless? Tried making few test prompts but ngl it's quite hard to tell. Apologies for highjacking the comments section, I just thought my question isn't worth a dedicated thread.
Merge model like black box, it contains lot of concepts.
I don't know what it can do, only test each tags step by step.
This specie/artist list from old FluffyRock csv tags.
I think is enough to cover most e621 concepts.
Details
Files
yiffymix_v42.yaml
Mirrors
yifurr_sunrise.yaml
yiffyfluffybutts_V061.yaml
yiffymix_v43.yaml
9thTail_juicyV03.yaml
indigoFurryMix_se01Vpred.yaml
yifurr_c3.yaml
fluffyrockanimedilut_v10.yaml
furationBETA_rebootedVPREDPartF.yaml
SD15ColorsplashV_v01.yaml
SD15ColorsplashV_v013.yaml
easyfluff_5000DefaultvaeFFR3CS2.yaml
subfurryanalog_v10.yaml
subfurryanalog_v20.yaml
subfurryanalog_v22.yaml
beanv6flufusv2_v10.yaml
9thTail_mainV02.yaml
9thTail_mainV01.yaml
9thTail_mainV03.yaml
pegasusmix_VPredV1.yaml
9thTail_altV03.yaml
steinmix_charlie2.yaml
steinmix_alpha6.yaml
yifurr_cid4.yaml
steinmix_alpha5.yaml
steinmix_alpha3.yaml
yiffymix_v41.yaml
indigoFurryMix_se02Vpred.yaml
yifurr_cid.yaml
steinmix_alpha6bL.yaml
fluffyrockE6LAION_e98E6laionE47.yaml
yiffymix_v44.yaml
pegasusmix_VPredV2.yaml
SD15ColorsplashV_v011.yaml
furationBETA_rebootedVPDModelbaseF.yaml
furationBETA_rebootedREFINEDVPREDF.yaml
fluffyrock_e184VpredE157.yaml
fluffyrock_e257VpredE11.yaml
yiffymix_v40.yaml
yifurr_2.yaml
yifurr_sun.yaml
yifurr_newmoone2.yaml
fluffyrockUnleashed_baseV10.yaml
9thTail_softV01.yaml
SD15ColorsplashV_v012.yaml
fluffyrock_e233VpredE206.yaml
yiffymix_v42.yaml
9thTail_softV03.yaml