LORA CHARACTER FROM FINAL FANTASY
Trigger Words: Tifa Lockhart
Trigger Words Optionnal: Main Outfit
Weight: 0.8
You can also take a look on what i generate on: https://www.pixiv.net/en/users/73135826
KOFI PAGE IS NOW OPEN FOR PRIVATE REQUESTS: https://ko-fi.com/judgenaegi
Description
FAQ
Comments (7)
Using all the same 'stuff' (including checkpoint) with the exception of your specific 'euler-a' (using the default "Euler A" and even transferring the data from "PNG Info" I am getting results that are the same.. except parts of the body are blurry. Sometimes the legs down will be 100% fine but the face will look like its out of the original silent hill 2.
Not sure what the issue is. I also use Forge.
can't say, im using comfyui on my side, everything works perfectly with many poses
@JudgeNaegi Hmm, I'll try more tests with tweaks. I have no idea why, your other loras work fine when i try them.
@WorstAirtist asked some friends of mine to give some tests on tifa and it's works perfectly on theyr side
Just posted what I get with a Hi Res Fix, and without (in that order) in either case I need to sleep. Maybe they can post one of their example images. I've never had this issue with one of your models before, which is why I'm confused my results are so poor.
@WorstAirtist this comfyui workflow he uses have inside some upgraded "hiresfix", it not presented in forge(as i know), but base hiresfix can create good enought pictures. "upgraded hiresfix" is 3 steps of hiresfix with add lightning lora to decrease generation time. (multistep upscale can fix some problems like non right finger count with better chances that 1 step, and if you add on this step some noise(comfyui can) you add some details too)
On forge you can use hiresfix x2(or x3 if you have good GPU) + adetailer on face with something like https://civitai.com/models/293448?modelVersionId=577989 this model (link on version, not latest one) for get ~same result
@randomizer89 I will look into it soon. Thank you.