Lora-scale: 0.7-0.9
Trigger Words: liz cather, long hair, black hair, green eyes
If you'd like to support my work, you can do so through Ko-fi!
Description
FAQ
Comments (11)
Poor results
Thanks for using my loras, Would you mind explaining more clearly?
@duongve13112002 You can try not to add the inscription of the clothing on the prompt word, or even "1girl". You will find that its compatibility is very poor on most models.
@MuMeGosui I don't quite understand your point. If you don't provide a description, how can the model generate an image for you? If it can do so, the model would have very poor creativity.
@duongve13112002 By the way, if my remarks feel malicious, please don’t get me wrong, because I used GOOGLE machine translation, so I am not sure that such content does not conform to the formal grammar. I have no malicious intent, thank you, thank you lora.
@duongve13112002 For example, if you change the costume and don't even add the prompt word "girl", the entire character will be very different from lora itself. You can try it yourself with other pony models, and you will understand.
@duongve13112002 In addition, this lora seems to have no effect on the Illustrious model.
@MuMeGosui This Lora works well with the Pony model, but it wasn't designed for the Illustrious model. Moreover, as I mentioned, you need to use multiple prompt phrases to create the character, not just rely on the trigger prompt. For example, without the prompt 'girl,' how can the model understand that it's generating a female character? This will result in poor-quality images.
@duongve13112002 I think you should try it yourself instead of focusing on "girl", because even if you use the prompt word girl, it is still tied to the clothing description. If you don't use the clothing description you specified, it won't work. After making this character, I think he has lost the existence value of lora itself. Of course, if you want to continue to maintain such quality, you are free~ I am only describing the results of use, just be happy with it♥
For what it's worth @duongve13112002, I have no idea what he's talking about. Downloaded the model, put your prompt words in (minus the clothes) and it generated whatever I wanted just fine. "Skill issue" perhaps? 🤷
@Vortigraph I think the issue is that he/she used pony LORAs for a model with the Illustrious architecture, which results in poor image generation. Currently, I’m waiting for this series to finish so I can update a new version for pony along with the Illustrious model.