This is my second style lora so go easy on me with this one.
V2 is a bit of a different beast to V1. Its a lot stronger trained so it matches the style and characters better, but at the drawback of being a littler over trained on the characters themselves. So you might see a little bit of Aleks' and Damian's features in all your characters, especially the hair. If that's a turn off, either turn the weight down a little, add some tags to your negatives, or use V1 which is still a decent LoRA.
For V2 I also recommend maybe experimenting with a lower weight of around 0.8 if you start getting messy backgrounds etc.
Always been a fan of this animator so it was quite fun going through all his animations to get the dataset.
Trained on 3 of his main characters listed below, along with the tags I recommend:
Aleks: aleks, rottweiler, green eyes, brown body, floppy ears
Damian: damian, anubian jackal, blue eyes, black fur
Amicus: amicus, wolf, grey fur, blue eyes, facial marking
Description
Cleaned up and retagged dataset. Added some images from their latest animations.
Trained on 149 images:
{
"engine": "kohya",
"unetLR": 1,
"clipSkip": 2,
"loraType": "lora",
"keepTokens": 1,
"networkDim": 16,
"numRepeats": 4,
"resolution": 1024,
"lrScheduler": "cosine",
"minSnrGamma": 5,
"noiseOffset": 0.1,
"targetSteps": 2980,
"enableBucket": true,
"networkAlpha": 8,
"optimizerType": "Prodigy",
"textEncoderLR": 1,
"maxTrainEpochs": 20,
"shuffleCaption": true,
"trainBatchSize": 4,
"flipAugmentation": false,
"lrSchedulerNumCycles": 3
}