Share your images here and leave a like if you enjoyed it! Get exclusive NSFW models at DeviantArt or Booth.pm.
LoRA based on the style of the artist Tando. Follow the artist and support if you like their work. No trigger word required. The end result depends on the base model used and the preferred weight. You can probably get a better adherence to the style by removing most if not all pony scoring/quality tags.
Recommended quality tags: score_9, score_8_up, score_7_up, score_6_up. Optional: source_anime
Recommended negative tags: score_4, score_5. Optional: 3d, lips
Recommended checkpoint: AutismMix_confetti
Description
FAQ
Comments (5)
Actually score_8_up is linked to realistic lighting so it might give better results without it. Just letting you know in case you didn't know already :)
Do you mean both score_9 and score_8_up? And I assume you're saying to remove that one? Because if it's linked to realistic lighting it's not ideal for flat anime styles then.
I spend too much time training and too little on inference to know better :D
Just tested without any of pony's typical quality/scoring tags, including the negative ones, and seems to work better. Thanks for the heads up!
@psoft No problem, and the realistic texture/lighting is related to score_8_up. score_9 is fine
Is there a place where this kind of thing has been documented? This is the first I've heard of this, but training is so idiosyncratic that I wouldn't be surprised if these weird biases were hiding in every corner (especially given the hiccups in Pony V6's training).
If the other score tags have unintended side effects beyond "make more good", I'd be very curious to know what they are!
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.