Thank you! ;-)

Yes, I will prepare some tutorial soon. Thnx for asking :)

@nodis46916414 that's a great compliment :-)

i will be doing that tutorial over the weekend, will most likely use some new model as a show case :)

@elric29200193 perhaps you can share some of your nice results on her model page for us? :-)

@elric29200193 hey hey, unfortunately, nothing yet but I'm doing a guide right now so I will have to figure something out :)

nice that you moved to a better GFX, you will get hooked on generations like I did (or you have already :P)

cheers!

@uzer i'm still doing the tutorial, i overestimated how much time it would take me to do it, I guess I'm in the middle of it :)

but to answer now -> yes, I still do dreambooth and then convert to lycoris

it is difficult for me to compare this way to the regular training of lora because I haven't done the regular one - I've managed to set this up before doing regular trainings and once I've had it set up - I didn't need to go to regular (I wanted just to compare but I figured I would need to tweak the regular one to the best params in order to make a fair comparison - and I never had time to do it)

 I am also interested in how to obtain good training results for real life faces.

I cover that part extensively in my tutorial. Perhaps I will be able to finish it tomorrow (Wednesday the latest)

@nodis46916414 i have made an official guide of my process ;-)

i will include that in the guide that i will be doing over weekend :-)

@HollowBrain exactly!

@shrrgsa play close attention to your A1111 console - there will be info telling you if the lora is missing (which means the name is wrong)

and also make sure that the LyCORIS extension is loaded, this is no ordinary LORA!

good luck!

this is something that still baffles me, I certainly need the lycoris extension, without it they won't work

in the console it does say "locon load lora method" so is recognized as locon

if you see samples made by other people - some of them have "<loco:" in their prompts - I imagine this is based on the configuration of these extensions, would probably need to see the code to check why some people need to use loco and some lora

currently, I believe it is because it checks whether "LyCORIS" extension exists, I don't have it so it probably defaults to <lora:

but it is still confusing why this change...

@PennyBoog I still have older controlnet but I will update and grab the latest models and will check

question: by "it does not work" - do you mean the image is generated but without the data from the lora/lycoris model?

@PennyBoog interesting indeed; can you try grabbing lycoris from another person and see if it is the same issue? and then try doing it with LORA, we would be able to pinpoint the issue this way perhaps

have you updated a1111 recently? someone else mentioned that a1111 broke something with lycoris, perhaps that is the issue and you just need to revert to the previous version for the time being?

I will update to latest a1111 later today and will investigate

I've written her down as TODO. If you are able to provide me a images for her dataset - this would be wonderful :)

@Dashdashie thank you, i've updated the spreadsheet:

https://docs.google.com/spreadsheets/d/1fOa6Jp_32DPM4Dc-oG2ILJOs0Xn1BydgbK2fbkp9k-c/

I have added it to the list but can't promise anything fast. If you really want her version - you could use my buymeacoffee, then I'll make extra effort to make that version for sure :)