Request
Hi, thank you for all the work you’ve put into this project. I really appreciate it. Is there a way to request custom loras or suggest some ?
You can give your suggestions I will see
Thanks for the loras but they are to rigid, i suggest describing the face a little less... A simple "a headshot of " should always generate a 100% accurate result... I am no expert but what i gather the training prompt should only have stuff that you need to remove or are temporarily like headphones or glasses.. and something like "almond shaped eye" you shouldn't put in the training prompts as its a core feature of the face.... Hope this helps you make better lora.. waiting for the next drop ☺️
@Hadeng
Noted it for the next batch
@juserisesseries
Yeah, I mostly use detailed captions to train much larger loras, to produce wider range of effect "style", and used the same template to caption these faces, it's ok to produce likeness even if abit diluted, but can be more grounded with a simple trigger it's true, I could maybe retrain some and see if it's worth reupload them
Here are a few loras I’d appreciate if you happened to train: Angèle, Pokimane, Léna Situations, Agathe Auproux, Meryl.bie. Of course, only if you want to train any of them. Also do you plan on open-sourcing the pipeline or training data ?
If you still take reguests ı would like to see ExtrÂEmily, MÂyaHiga, Em1ru,MÂlenaTudî thanks in advance
@zdsfegbhn
I've added "Pokimane, Léna Mahfouf" in the new batch of 50. As for the pipeline its mostly hermes skills that rely on external dependencies like comfyUI, unsloth studio, runpod, but If you want a synthetised tasks order to reproduce it I can share it.
@darkdays341
I note, idk If I will run a third batch tho
I know you may not add any other characters, but in case you do, it'd be nice to have some HP characters - the golden trio, Voldemort, Dumbledore, Draco etc. I'd especially love a Snape lora