This is initial test for tifa_build
.. still building the database/models. will be updated soon
this mk.01 -> result is still unstable / inconsistense, should use parameter (below 0.5 for better result) ex : <lora:tifa:0.48>, or maybe just try <lora:tifa:0.68>. this mk.02 -> improved on accuracy. should use parameter (below 0.7 for better result)
better compatibility with "perfect world" models https://civarchive.com/models/8281/perfect-world
check my twitter here for the art result and more showcase XoXo (@ArSphere) / Twitter
test _comparison sample. on mk.02
prompt : tifa, final fantasy 7 remake, (mature female), milf, shade of brown hair, wind. very white skin,pale skin, masterpiece, ultra realistic 8k cg, very long hair, beautiful , , reddish brown eyes, huge breasts, left only earpiece, stick earpiece, red earpiece, {one earpiece}, left earpiece, shiny skin, clean, beautiful face, pure face, pale skin, long hair, (huge breasts, beautiful breasts, ), shiny skin, skindentation, (((((black swimsuit, ))))),
sampling : DPM++2M Karras
Seed : 2436795641
resolution : 1024x1400
Models : perfectWorld_Baked_VAE
Description
-improved accuracy,
-best use on half body and potrait
FAQ
Comments (13)
where is the lora-mixTifaFF, can`t find it,any one know?
i change it to match the file and the sample image (because i use the same name in the sample, didnt want for people to keep asking me about the lora name/which lora). seems after maintenance yesterday, now, civitai downloaded file name is programmed to match similiar to the title. title=file name
Not able to find your model. When I use the model in the description, Perfect World, I get the hash: 79e42fb744 and it's different from perfectWorld_perfectWorldBakedVAE - TIFA (ca883fecc7). Where can I find the model used in the examples?
make sure to add my lora. its perfect world baked VAE+tifaArsphere_mk02 (set at 0.5 instead 1) should be get the same result. i will send the screenshot. check review
Can I ask, what is the word "fuze" mean in the negative prompt? is it an embedding? I only ask because it has high weighting in your images, and im wondering why I can't perfectly reproduce your results
tbh about negative prompt i just randomly copy from other people and not really giving antention to it (at least for now). because most of the time my result is good without tinkering the negative
Why the fingers always has display problem?each images has more than 5 fingers
yeah, the AI is still imperfect. but soon maybe.. this AI technology is moves realy fast
Thank you ArSphere. The best Tifa lora here, but I'm having a very hard time generating a male body. Can anyone suggest a compatible male lora?
hi, late respon, been busy lately.. about male, i am sorry i dont know, i still havent really checking about it. but yeah seems nice if we can add cloud there
Question for you. Been using these alot and definitely think both are great but I do think Mk2 has a better face generally. However, for some reason mk2 seems to give a very warm/brownish tint to everything compared Mk1? Is there someway to counteract that you can think of?
Thank you for the fantastic tools either way.
i usually use body fix before adding this lora, body fix strength should not be high, maybe around 3 or 5 or else it would dilute the result. alternatively use inpaint
She's beautiful, Tifa and Cloud model is posible? cloti♥♥♥ thanks
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.