Loading
Du hast einen ComfyUI-Workflow? Erstelle direkt im Browser eine verständliche Workflow-Dokumentation.Workflow verständlich machen.Jetzt ausprobieren ↗
Hi there! This is the Anima version. Don't expect too much from this model—I haven't been able to get the hang of training ANIMA LoRAs yet. I'm not sure if my dataset is the issue or if something in the training...
Modellquelle
Quellenbeschreibung
Hi there! This is the Anima version. Don't expect too much from this model—I haven't been able to get the hang of training ANIMA LoRAs yet. I'm not sure if my dataset is the issue or if something in the training settings is off, but I wanted to post this one since I managed to get a few decent results. If another LoRA comes out for this character, feel free to switch to it, and if you manage to get any interesting results, let me know!
Training Settings & Details:
Images: 29
Batch size: 1
Steps per epoch: 29
Total epochs: 50
Total steps: 1,450
Quellen
1 QuelleModellartefakte
1 ArtefaktLolaLoud(Character)[Illustrious].safetensors
safetensors · 218 MB · SHA-256 7672ade3803e…8348 · Civitai
HerunterladenQuellenauszüge
1 AuszugSource context: 558 downloads · Type LORA · Base model Anima
Resolution: 1024x1024 (automatic bucketing)
Network dim: 8
Network alpha: 8
Learning rate: 5e-5
Optimizer: AdamW8bit
Scheduler: cosine_with_restarts (3 cycles, 100 warmup steps)
Precision: bf16
Final loss: 0.0563
Training time: 2h 20m (~5.8 s/step)
Hey! This time I have a character from The Loud House series, one of my favorite characters, Lola Loud. I think the model turned out a bit overcooked, but it’s been giving me pretty good results. The ideal weight is between 0.8 and 1—I’m not responsible for anything that happens outside those limits 🤣.
How to use: Don’t forget to add the tag lola loud . Use the images I provided as examples, and include the tags style parody , parody to get a style very similar to the show, or official style to stick to the official look!
