Like my models? Support the cause.
Update: V2
Trained with Epicrealism. For best results i do recommend this checkpoint. Should work with other though.
The model now know 2 different variants of the concept via the tags:
thinyoke
thickyoke
seems to be consistently producing consistent results with either of those.
V1
Seems to work ok with the checkpoints i have tested.
Indfluence between .8 and 1.
Trained with basesd 1.5
Some posssible reeinforcement tags: hands up, arms up, bdsm, bondage etc
Description
FAQ
Comments (9)
could you please post some reproducable exampleimages with complete prompts? everything my webui produces is utter crap, even outrun by anime-loras :(
Thank you very much!
the second image in the gallery has complete prompt. the lora seemed to work pretty well with my testing: porompting the triggerword with minimal additions should work. not sure why it doesnt for you.
@Joschek then if its not in the prompt... what is your native resolution? 512*512? 1024*1024maybe? or do you upscale?
@maybellesunfor those its like 512 by 512 and then upscaled, but i remember triyng different aspect ratios and resolutions and it seemed to have no major problems
@maybellesun wait a second... i just tried with the civitai online generator and it does seems to have bad results here, maybe i uploaded the wrong file. ill try to get that right asap
@maybellesun the file was correct, but i found possibly what was going on: adding the (for epicrealism recommended) negative prompt : "cartoon, painting, illustration, (worst quality, low quality, normal quality:2)" returned it back to good resultts in the online generator here. let me know if that fixes it for you too... i always forget that negative prompts do actually play some role
@Joschek thank you for your enthusiasm. I tried with negative and without already, tomorrow i will post some examples to show you where I struggle. Good night!
@Joschek Well, it seems i have mixed up several parameters... and that only when they come together correctly it goes out nicely... i changed them all yesterday but did it maybelle-style... sorry to bother you and thank you for your work! Now that i know where to look at maybe i can use your other ones as well!
(could you maybe make a lora of bondage- mittens?)
CFG: 5 (i usually bolt it at 7...)
Upscale Latent (i usually do not use it because my machine is powered by potatoes)
And definitely not a seed of 0.
Stupid Maybelle! (the picture i posted is not meant to be artistic, but i am afraid the site ate the "thank you" comment sticking to it....)
Thank you very much,
Sincerely,
Maybelle
I am using your model well in comfyui.
With Lora Block weight
[1,0,0,0,0,0,0,1,1.2,0.8,1,0,0,0,0,0,0,0,0,0]
as the block weight.
In addition to the default tags, we use the following tags.
[positive]: "arms side by side"
[negative]: "hands up, arms behind back"
Another suggestion is to use together with
another Yoke Lora model by "TrinityDeVayne".
[https://civitai.com/models/96043/bondage-metal-yoke-front]
In this case
"Joschek's model" would have [1,0,0,0,0,0,0,0,1,1.2,0.8,1,0,0,0,0,0,0,0,0,0] for the
For "TrinityDeVayne's model" use [1,0,0,0,0,0.6,1,0.6,1,1,1,1,1,0,0,0,0,0,0,0].
weight is less than 1.5 for both models combined.
When used together add tag to
[negative] : "wearing leather mittens, gloves"
and adjusting weight accordingly,
It is also possible to exclude mittens