Hi Malcolm, just asking to confirm...if you're releasing a lora of a person you've done a lycoris for already, is it meant to be an update using your latest techniques? In my comparison here using the same seeds+prompts for example, the new lora is more neutral and 'gentler' while the older lycoris leans a bit harder on her features. I know the general rule is to use whichever you end up preferring but I'd like to know your intentions :)

Since you are taking requests, how about Jane Levy.

I tried following a guide to create a TI myself, but ran into the "AssertionError: No inf checks were recorded for this optimizer." error. Have a mess of ~500 img, selected a few dozen and edited out watermarks already. Don't know what I'm doing yet, but they might help in creating a LyCO/LoRA. If they are useful, let me know where to send them.

While you're taking suggestions, may I humbly ask for Isabella Merced (she was in the Dora the Explorer movie, and Transformers Last Knight), and Bryce Dallas Howard (more specifically, the version from the first Jurassic World with the straight short hair). Oh, and Alanah Pearce would be nice too (the Aussie video game reviewer chick). Thank you, and great work.

I think Julie Bowen is a challenge, since she has some unique features that seem to cause haywire when you try to render her. Katie Holmes is a similar case, and they coincidentally are my two holy grails.

Am I using these LoRA/Lycoris models wrong, or is it expected that the base model will know better than this LoRA how a celebrity looks? I'm getting better result by just using their name in the base model is what I mean

Why is the image I generated after downloading not at all similar to the model?

this lora is non-functioning, check your file

Man, this thing is about as flexible as a League of Legends player. Unless it's a close-up shot, there is no resemblance whatsoever.

hi, thank you for the taylor swift lora! suggestion for a Lora: Rebecca Mir, she is a german model and tv moderator, thank you