This model is a LoCon. You MUST install the Lycoris extension for it to load.
I'm using Lora Block Weight. I believe you can also use Additional Networks and SD Webui Lycoris.
UPDATE 4/27/2023: I've hit a training plateau so I'm in the process of adding a bunch more images to the dataset, including some more complicated stuff like intertwined fingers. I'm probably going to have to drop the learning rate some more, so things may be slower from here on. I'll keep everyone posted as things progress.
UPDATE: Prompting advice for beta 2:
This is a completely new train on top of vanilla Stable Diffusion 1.5. I did this based on the advice of a fellow enthusiast, and it's surprising how much more compatible it is with different model. It doesn't mess with the style of your model at all as far as I can tell, and it really only affects hands and occasionally arms, leaving everything else untouched.
It seems to work best at a strength of 1, although turning up higher than that (1.5, 2, etc) can help it on some images at the cost of making it worse on others. No need to mess with your CFG scale, as it doesn't cause things to overcook at these levels.
Freely mix it with other LoRAs.
I've had best results putting "nice hands, perfect hands" in the positive prompt (increasing the weight makes things worse), and "(extra fingers, deformed hands, polydactyl:1.5)" in the negative prompt. This is on EnvyMix v1 (and probably RevAnimated), but YMMV for other models.
"Bad hands" negative embeddings appear to make it worse, although I haven't tested this extensively.
As usual, this won't work miracles, but I do find that over a large number of images, it does make things generally better on average. Hopefully this will continue to improve with a few more nights of training.
Prompting advice for alpha 3 and beta 1:
Note that this advice is for RevAnimated 1.2. YMMV with other models.
It overcooks things a bit, but you need the strength set to 1.0 for it to really work well. You can work around this by reducing the CFG value to 5 or 6 or so. I've had good luck with enabling the dynamic thresholding extension and setting it to mimic CFG 5, and then I can set my CFG value to 9 or 10 and things come out fine.
I tried using it with another LoRA and got some pretty strange results, so YMMV there as well. Right now I'm just trying to get it to work consistently in a simple use case.
Oddly, I think it's regressed a bit on hands in neutral positions, but it's noticeably better at more complicated interactions, such as holding objects (which is why I have so many pictures of blacksmiths and librarians in the example images).
Keep your prompts simple and it tends to do better.
With RevAnimated, I tend to get 1 or 2 usable images out of every 8, with a bunch of other ones that are pretty close and can probably be fixed with inpainting.
Prompting advice for alpha 2:
It's getting stronger now, and it works best around strength 1. Setting it to 1.3 like the previous version will make things look bad.
My negative prompt is still "(extra fingers, deformed hands:1.15), (worst quality, low quality, poor quality, bad quality:1.35)"
I had good luck just putting "nice hands" in the main prompt.
Prompting advice For alpha 1:
Your prompt should contain these words: "beautiful hands, perfect hands, fingernails". I've had the best luck with them towards the middle, and at no emphasis.
The alpha1 LoCon seems to work best at a strength of around 1.3 (on RevAnimated 1.1, where I'm testing it right now -- YMMV for other models)
Don't use negative embeddings for improving hands. When I removed badhandv4 from my negative prompt, things improved noticeably. You may want to try without any negative embeddings at all. I haven't used them for a while now.
My negative prompt is: "(extra fingers, deformed hands:1.15), (worst quality, low quality, poor quality, bad quality:1.35)", which I arrived at through a lot of experimentation adjusting strengths and terms one at a time. It should work decently well.
This all gives me hope that there's a real shot at solving hands on SD 1.5. Even with good prompting, I'm generally not getting perfect results, but things are close. I'll consider this done when it creates well-formed hands without having to add anything to the positive or negative prompt.
Now back to your regularly scheduled readme...
I'm testing the theory that maybe the reason MidJourney's hands are so much better now is that they just took the time to specifically train a network on a high quality set of pictures of hands, and literally nobody else has actually tried. This LoRA definitely isn't at MidJourney levels yet, but I've been training it over night for several nights now and adding to the dataset where it appears deficient, and quality seems to be steadily improving. As such, I'm going to post this now so people can start using it. Consider this an early alpha -- I'll only stop updating once it stops getting better.
Example images are cherry-picked. Please don't expect this model to make all of your hand generations better. It may even make some of them worse, so you should evaluate its usefulness on a large number of images and not just one. If it works for you like it does for me, the a lot of your results should be the same or better quality (some will just be bad in different ways).
Description
Restarted training from scratch, because apparently training on vanilla 1.5 is actually better in terms of making models that don't overcook things or change the style very much. This new version is still in need of more training, so it's not quite as effective as the old one, but it does seem to, on average, improve things a bit, and it works across a lot more models and doesn't mess with style at all, so I think this is probably the right direction to go in. I'll play around with prompting a bit and update the main description with advice.
FAQ
Comments (112)
I get this when I use beta2 :
ctivating extra network lora with arguments [<modules.extra_networks.ExtraNetworkParams object at 0x000001F27F74AF20>]: RuntimeError Traceback (most recent call last): File "E:\Ai\stable-diffusion-webui\modules\extra_networks.py", line 75, in activate extra_network.activate(p, extra_network_args) File "E:\Ai\stable-diffusion-webui\extensions-builtin\Lora\extra_networks_lora.py", line 23, in activate lora.load_loras(names, multipliers) File "E:\Ai\stable-diffusion-webui\extensions\a1111-sd-webui-lycoris\..\..\extensions-builtin/Lora\lora.py", line 214, in load_loras lora = load_lora(name, lora_on_disk.filename) File "E:\Ai\stable-diffusion-webui\extensions\a1111-sd-webui-lycoris\..\..\extensions-builtin/Lora\lora.py", line 176, in load_lora module.weight.copy_(weight) RuntimeError: output with shape [1, 320, 1, 1] doesn't match the broadcast shape [1, 320, 3, 3]
i had the same error, i dont know if the problem is why i was using some other lora together
I re-uploaded it. I don't know how the upload from 4 hours ago even got published. I didn't do it deliberately.
Download it again and try. Also, make sure you have the Lycoris extension installed, and that you stop and restart automatic1111 after you install it. If it still doesn't work, let me know and I'll delete the entire version and upload it yet again (it's working just fine for me).
@_Envy_ pls how i install lycoris extension
@_Envy_ Thanks, im gonna test again
@_Envy_ I was able to get it to work via additional networks, with it placed in that lora folder. If I try to run it as a typical lora via <lora:envybetterhandsLocon_beta2:1> I get the same error.
Okay, I switched to the vlad UI recently and I was apparently confused as to what plugin was doing what. To use it the way I'm using it in my prompts, you'll want this one here:
https://github.com/hako-mikan/sd-webui-lora-block-weight
It hijacks the built in lora prompt system so you can also use it for LoHAs and LoCons.
hi, did you use dynamic-thresholding extension for beta2?
this looks very promisinng. def gonna keep my eye on it. Hopefully your theory proves true.
you are great,pls continue this work,stabl diffusion should better than midjourney !!!!!
pls can you help me how i can install it im new here pls
you download the file and then paste it in this folder <<the folder u initialized ur SD in>>\stable-diffusion-webui\models\Lora (i use the webui but it should still be the same if u use the CMD SD)
I keep getting this error:
activating extra network lora with arguments [<modules.extra_networks.ExtraNetworkParams object at 0x000002ABDC0B6410>]: RuntimeError Traceback (most recent call last): File "E:\webuiA1\stable-diffusion-webui\modules\extra_networks.py", line 75, in activate extra_network.activate(p, extra_network_args) File "E:\webuiA1\stable-diffusion-webui\extensions-builtin\Lora\extra_networks_lora.py", line 23, in activate lora.load_loras(names, multipliers) File "E:\webuiA1\stable-diffusion-webui\extensions\a1111-sd-webui-lycoris\..\..\extensions-builtin/Lora\lora.py", line 214, in load_loras lora = load_lora(name, lora_on_disk.filename) File "E:\webuiA1\stable-diffusion-webui\extensions\a1111-sd-webui-lycoris\..\..\extensions-builtin/Lora\lora.py", line 176, in load_lora module.weight.copy_(weight) RuntimeError: output with shape [1, 320, 1, 1] doesn't match the broadcast shape [1, 320, 3, 3]
I have lycoris extension installed and the file in models/lora folder. Please help.
EDIT: Also have Lora block weight installed.
Try not using Extra Networks, and instead just specifying it in your prompt like <lora:BetterHands-beta2:1>. Other people have had luck with Extra Networks, but Lora Block Weight makes it work the same way as other loras do in your prompt. Let me know how it works, and send me a screenshot if you still have trouble.
@Maple_Senpai @_Envy_ As far as I know, KohakuBlueleaf's extension is the only one to add Lycoris support. If you've got it in the models/Lora folder, it's in the wrong spot. It should be in a folder underneath extensions. You should either install it through the webui -- go to the extensions page and tell it to install from the URL of the extension -- or, if you're at home on the command line/terminal, you can just do a git checkout inside the extensions folder.
When I've tried to use a Locon in the past inside the Additional Networks extension, I got the same error (different tensor sizes), even though I had the Lycoris extension installed. You can only use a Lycoris (LoCon/LoHA) by adding it to the prompt with <lora:name:weight> and not through the Additional Networks extension.
I hope that helps you resolve your issue.
Why do you tag it as LORA, when it is a Lycoris LoCon?
I tagged it incorrectly, and now I'm worried that if I fix it, people who know about it will search for it again and think it disappeared.
I think this has a tendency to stick things into the subject's hands
I installed goodhands_beta2.safetensors to lycoris folder, and gives me error..
reading lora E:\stable-diffusion-webui-master\models\LyCORIS\GoodHands-beta2.safetensors: AttributeError
Traceback (most recent call last):
File "E:\stable-diffusion-webui-master\extensions\a1111-sd-webui-lycoris\lycoris.py", line 152, in __init__
self.metadata = sd_models.read_metadata_from_safetensors(filename)
AttributeError: module 'modules.sd_models' has no attribute 'read_metadata_from_safetensors'
is it nsfw supported ? such as for example,penis grab teaser etc
I haven't tested it with anything hardcore, but considering that it doesn't change much other than hands, it'll almost certainly work fine.
what is the trigger word
There's no specific trigger word, but I recommend reading the prompting advice in the description.
@_Envy_ btw i get this error using it ERROR: closing handshake failed
I don't know what that error is. Is the program crashing, or does it generate images?
You stated the train was with suggested settings/ideas from others- care to share the love a bit on what settings you found worked for you? I've been finding locon/lora/loha very hit and miss at times, or changes the styles too much. Thanks!
Mainly that a lora/locon will mess with the style less and be more compatible with multiple checkpoints overall if you train it on top of vanilla stable diffusion 1.5. I was mostly intending this for anime, so I was training it on anime-related checkpoints, so I was pretty surprised to find that it worked way better trained on top of SD1.5, even on the checkpoint I was originally training it on.
So yeah, if you want to avoid it messing with the style, base it on SD1.5.
can i use easynegative in my negative prompt with this?
I haven't tested it, but probably. The best way to experiment is run like 30 randomly generated prompts with it enabled, then those 30 again without it, and compare which is better. That's how I've been determining whether it's working.
@_Envy_ thank you, this lora really works, I couldn't avoid generating with this now... the chance i get usable image has improved a lot. are you planning release better version in the future?
Yes, but I've hit a wall with these training settings and I need to do some fairly heavy reading about what they all mean so that I can come up with something better and start over.
So is it LORA or LICO? When you download via CIVITAI HELPER, it downloads to the LORA folder. And your examples say <lora:GoodHands-beta2:1>. But you get errors in the console. It works only when <lyco:GoodHands-beta2:1>.
What for?
Unfortunately there are multiple modules that implement LoRAs, and I'm using one called Lora Block Weights, which keeps loras and lycos in the same directory. I'm afraid to move this now because I don't want people to think it's gone. Using it as a lyco is fine, though. It'll work just as well that way.
我发现的问题是用了将hand放在正向提示词里面，画面会突出手，而很多时候我并不希望手出现在画面主要位置，只是希望它出现的时候别那么奇怪→ _ →
perhaps you can try lower the weight or put the prompt in the last, help a little
Is it a LoCon or A Lora.Cause litle says LoCon but the description says lora
same here... idk where should I put the file
They are the same, they are installed and used the exact same way but for LoCon you will need to install this extension: GitHub - KohakuBlueleaf/a1111-sd-webui-lycoris: An extension for stable-diffusion-webui to load lycoris models.
Just install it just in case you need it.
@OdoxiAI I have "a1111-sd-webui-lycoris" and "a1111-sd-webui-lycoris-main" installed, where do I have to put this locon/lora file now to make it work? I can't find the tutorial mentioned in other comments
really need this!!thank you!
My lora have stopped working. yesterday they were working fine now not anymore. what can i do?
The locon extension is currently broken
https://github.com/KohakuBlueleaf/a1111-sd-webui-locon/issues/40
\stable-diffusion-webui\extensions\a1111-sd-webui-locon\scripts,open (main.py)
line 371: lora.mtime = os.path.getmtime(lora_on_disk)
line 373: sd = sd_models.read_state_dict(lora_on_disk)
@atbgams fixed thx
can i set this up on comfyui?
Hello, I have a question. When I run this Lora, I encounter the following error.
loading Lora C:\Users\JJAY219C\stable-diffusion-webui-master\stable-diffusion-webui-master\models\Lora\GoodHands-beta2.safetensors: RuntimeError Traceback (most recent call last): File "C:\Users\JJAY219C\stable-diffusion-webui-master\stable-diffusion-webui-master\extensions-builtin\Lora\lora.py", line 215, in load_loras lora = load_lora(name, lora_on_disk.filename) File "C:\Users\JJAY219C\stable-diffusion-webui-master\stable-diffusion-webui-master\extensions-builtin\Lora\lora.py", line 176, in load_lora module.weight.copy_(weight) RuntimeError: output with shape [1, 320, 1, 1] doesn't match the broadcast shape [1, 320, 3, 3]
I tried deleting and re-downloading it, placing it in the Lora folder. I also downloaded Lycoris. However, I still face the same issue. I noticed that you mentioned this Lora is based on Stable Diffusion 1.5, but mine is updated. I believe I downloaded the latest version of Stable Diffusion around May, which is version 2.1. Could this be the reason for the problem? Because apart from this Lora, all my other Loras run without any issues. Could you please help me understand what might be causing this problem? I truly appreciate your assistance, as I really want to use your Lora.
Yes, 1.5 and 2.1 have a different architectures and the LoRAs from one aren't compatible with the other. You need 1.5, or any checkpoint on this site that's based on 1.5 (which is most of them).
Hi, in the examples this thing is very useful, but for me it doesn't work like LORA, no matter what priority I give it, I just keep generating bad hands. I downloaded the latest version of this lora in the folder (models/lora), SD version 1.5. Any options on how to make it work? (the only thing that has changed is that the hands on my generations are now glowing blue, but still with the wrong hand anatomy)
good job
My base is 1.5. When I load it with <lora:GoodHands-beta2:1>
It report error
loading Lora C:\novelai-aki-v2\models\Lora\GoodHands-beta2.safetensors: RuntimeError
......
RuntimeError: output with shape [1, 320, 1, 1] doesn't match the broadcast shape [1, 320, 3, 3]
Other lora module working fine. Does anyone can help me with this?
不是lora模型 仔细看下面说明
this is awesome, I hope someday it's gonna make the hands look beautiful. Waiting for the next update, thanks for this 😘
can someone explain to me as if I were an idiot how to use this please? or tell me where the detailed instructions are
thanks
if you didn't find out in the meantime, you need to install 2 extentions. the links for them are on the first phrase of the description, aditional networks and lycoris. Once you have them installed in your webui, you'll have a folder in your models for the lyco types such as this.
私も純粋に高品質で手の写真を写実する訓練があって、機会があればあなたと交流することができることを望みます
Silly question, should I put it in the "Lora" folder or in the "LyCORIS" folder?
LyCORIS. Use it with <lyco:> instead of <lora:>.
You're doing an amazing experiment here. Thanks a lot for sharing. Fingers crossed for the evolution of the Locon.
A new learning rate scheduling algorithm just dropped, so I'm going to try training it further in the next couple of days. Hopefully I'll get past the wall I hit earlier.
Full Instructions and results!!!:
TLDR; Accurate hands in 90% of images is possible with ReV Animated! :-D
A few months ago, using this required multiple extension installations, then it only required a single installation, and as of 19 August, 2023, A1111 supports LyCORIS natively. Here are instructions for using:
1. Make sure your version of A1111 is up to date. There are instructions here: https://aituts.com/update-stable-diffusion-webui/)
(There is no need to install any extensions to support LyCORIS.)
2. Download the safetensors file from this civitai repo and put it in stable-diffusion-webui/models/LyCORIS. The file should be called "GoodHands-beta2.safetensors". You will need the exact file name for the positive prompt.
3. In your positive prompt, add "<lora:GoodHands-beta2:1.0>". Technically, you may use "lora:" or "lyco:" at the beginning of the tag, but you must put lycos and locons in the "LyCORIS" directory, and you must put LoRAs in the "Lora" directory.
4. (Optional) In your negative prompt, add "(polydactyl:1.5)". This will not improve hand quality, but it will tend to keep hands out of the scene completely more often. (Prompting is a vast field though, please experiment and use the prompts you find work best.)
5. And finally! Click "generate". :-)
To compare your results (sanity test):
Reuse the same seed, prompt, and settings, but remove "<lyco:GoodHands-beta2:1.0>" from your positive prompt.
You will regenerate the same image, but the hands (and possibly arms) will be very slightly different, with the rest of the image untouched. (If you are using xformers, every image generation will change slightly even with identical settings--things like leaves, grass, hair, etc. will vary slightly.)
Results: ReV Animated already produces ~50% nearly perfect hands at 768*1024 resolution. This LyCORIS changes hands in most outputs, and its overall accuracy is ~65%.
However! ReV Animated's 50% accuracy and BetterHands' 65% are different sets! By cherry-picking between this LyCORIS and your base model, the overall accuracy for the exact same image rises to 90%!!!!
Pretty sure the current built-in lora and lyoris extension extension made these setup instructions obsolete months ago. Compsable LoRA is also known to hijack torch procssing at start and generate frequent runtime errors.
@nashidaran I will look into it. Thanks for the info.
Looked into and updated / fixed! Additional Networks can also do composable lora's encoder / unet splitting, you just have to check the checkbox for it. :-)
@nashidaran You are correct, Additional Networks can also separate the text encoder / unet weights by enabling an extra option. (Composable Lora is not "obsolete" exactly, and all extensions hijack Torch's loading process. It generates exactly 1 error in some cases by checking for a user option the wrong way, fixed in one line of code in an upcoming fork switch.) I have updated the instructions here to mention enabling separation.
@quintessentialfo4180 I'm not sure but maybe we can use this (separate te and unet in arguments) instead of Additional Networks... Or does it work differently?
@Bjorno Yes! I just tested by running multiple same batches with / without Additional Networks installed, and it works correctly. Negative separation works well using only the LyCORIS extension. ("Dark" in the positive prompt makes the image darker, "Dark" in the negative prompt makes the image lighter.) I will update the instructions, thank you! :-)
hi, i found that locon is deprecated and calls errors and not compatible with addnet, also kohaku says:"basically the best solution now is: install lycoris, disable built-in lora, move all your lora/locon/loha/lokr/ia3/dylora file into LyCORIS folder", so, can you write updated guide with final explanation about: delete locon OR return to last working version and turn off addnet; how to use addnet; is it will work if do how kohaku says; how to use lycoris arguments te unet dyn
@Maficks I have never used locon before, and the built-in LoRA is working without errors in my installation. Could you find a link to Kohaku saying this? Is this a video channel, a news site, or another commenter? I would like to see if I can replicate the problems they encountered. Thank you for your help! :-)
@quintessentialfo4180 https://github.com/KohakuBlueleaf/a1111-sd-webui-locon/issues/40#issuecomment-1560727403 there, kohaku the creator locon and lycoris webui extensions, but how you never used locon method lora loading if many loras on civitai created as locon
@Maficks Found the problem and updated the guide with (hopefully) correct instructions. Thanks again for your help. :-)
@wazaza3131 I haven't found prompts good at keeping hands out of the frame, although to be fair, I've barely looked for them. For SDXL, the prompt "hands behind back" failed in 100% of tests I ran. And for now, RevAnimated is still generating good hands a little more often than SDXL, although they're very close.
With the automatic 11111 v1.5 update, lycoris are now supported, but when I add them to the prompt they have the "lora" handler. Is it still necessary to change the handler to "lyco" or does the program know that it is a lycoris automatically?
@kukain Thank you! I tested it this morning, and I've updated the instructions here. :-)
Changes: 1) You no longer need the lycoris extension. 2) During prompting, you can use either "lora" or "lyco", they trigger the exact same function, so it makes no difference which you write. Whichever you use, A1111 will select the correct file.
However, this is unchanged: locons must be in the "models/LyCORIS" directory, and loras must be in the "models/Lora" directory. (If you are unsure if your Lora or Lyco is working, check your console for errors.)
Thanks again!
for me beta2 0.5 is the ideal weight. 1 changes too much. Success rate is still not that high but pretty helpful in some cases.
Thank you so, so, so, so much for this. This works a charm on various checkpoints. It takes playing around with the weight a bit to find the sweet spot, but it does the trick every single time.
Hey just wondering, I don't believe there is an extension to download for warpfusion, do you know of a way to get it to work in there?
Thank you for saving my day! After reading carefully your advices and some testing, it works as i wish. You´ve done a good job!
...Never working, not sure what happened
same
It's never worked for me either. I have the lycoris extension installed but no matter what I do I the the goodhands model doesn't make an impact on my gens.
@ZZZ_Bear @tokoat
I found something.
If I use Hires. fix, set Hires50+step and Hires0.7denoising strength
I will have 1/100~5/100 rate to get nice hands.
GL, guys.
@ricky61204 My fix was removing the additional networks add on which another person posted that the extension causes a conflict with the better hands lora.
@tokoat ...So now ur handsfix is working?
I only use this EnvyBetterHands LYCO.
So I still finding why my EnvyBetterHands is not working.
Is there a license for this LYcoris?
I would like to know about commercial use, etc.
the change is a little bit subtle. I'll have to play with it a bit more to be sure. still, i'm hoping for some positive changes.
AMAAZZING!!
Idk why, but it changes style to incredibly cartoony for me, for some reason. It's almost ms paint quality, lol. Yes, it fixes hands, but in this state, it's unusable for me.
You can inpaint it with this lora to fix the hand(s), and to de-cartoon them inpaint it again with a realistic checkpoint and without the lora but WITH ControlNet to keep the shape of the hand correct.
is this a positive or negative prompt locon
Positive.
Hi, I would like to know how to install this LYCORIS?
You can install by adding this URL to your Stable Extensions tab.
https://github.com/KohakuBlueleaf/LyCORIS
LoCon Loading is now built in on this extension, so to activate you just have to use the standard Lora loader from Stable
Some kind of update like regular lora.
@Guikof this is no longer needed, SD is now used as regular lora
found at official git: LyCORIS models are also supported in the following interfaces / online generation services (please help us complete the list!)
edit: as I see in Focuuus git discussion, also supported
My guess is that this doesn't work and is just an optical illusion caused by the 'perfect hands' and some negative prompts. I don't see a difference when I remove Lora (the lycoris), so we need more clear evidence that it works.
we dont need anything, just test it out yourself and see if it helps you, not that hard.
Somebody did a video on it. It's not a silver bullet; it just makes hands render correctly more often.
It has an optimum weight, and I have a feeling that some later models may have already merged it in (which is 100% fine and allowed by the license). In those cases, you may be completely correct -- it wouldn't do anything to models it's already merged into. So in short, it may not be doing anything. Or, you may need to do 50-100 generations before it has a statistically significant chance to prove itself. Either way, it's your choice whether to use it or not. I tested it extensively back when I created it (blind tested and compared 100+ images), and it helped back then.
SDXL version? Please 🥺 🙏
I've tried, but I haven't actually been able to make anything that helps.
@_Envy_ Ah well. It was a valiant effort.
nice model
nice model
Just posting this comment here to prove my identity to Yodayo.
Details
Files
GoodHands-beta2.safetensors
Mirrors
GoodHands.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
envybetterhandsLocon_beta2.safetensors
GoodHands-beta2.safetensors
EnvyBetterHands LoCon.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2_Envy_.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
Meinen.safetensors
GoodHands-beta2.safetensors
envybetterhandsLocon_beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2_Envy_.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
envybetterhandsLocon_beta2.safetensors
GoodHands-beta2.safetensors
lragudhend.safetensors
GoodHands-beta2.safetensors
a_probablybetterhands.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
好手.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
hands.safetensors
玄学修爪子GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
envybetterhandsLocon_beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
hand.safetensors
修手_GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
gudh.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2_2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
BetterHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors
GoodHands-beta2.safetensors