
Continuing with creating Loras of every (core) Rugrats character, the 2nd in the series is Chuckie Finister!
How this was trained
Lora was trained with the "AutismMix Confetti" model.
Just like my Angelica lora, Training data was redrawn from existing official art in 1024x1024 at 300 dpi. No fan art was used.
Training steps of 1200 and epoch of 2.
Recommended settings
I usually start a prompt with the following:
score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, source_cartoon, BREAK, rating_safe, BREAK ChuXL, a little redhead boy with messy hair, freckles, purple glasses, buck teeth, blue shirt with red collar, green shorts, yellow socks and untied red sneakers,
Lora was tested in ComfyUI. I can't say if this will all work in A1111. May need to do more testing with that since I don't use it that often these days.
A strength_model from .75-1 and strength clip of 2.00 should yeld show accurate generations. Recuding the clip to 1 and model below .50 should stick to the model/lora style.
I recommend pairing with a style lora and embedding.
A lot of models from 'Styles for Pony Diffusion V6 XL (Not Artists styles)" work best. Smooth style looks nice.
The Pony PDXL Negative Embeddings works well.
Some weirdness
Occassionally, the lora generates double pupils in Chuckie's glasses. Using a different resolution outside of 1024x1024 tends to make him look a bit older. If you want it that way, I recommend doing that.
Misc
Going forward, all future loras I create will be for PonyXL only. No more 1.5 and no more SDXL. Model is NOT available on the Civitai generation service as I do not own the rights to this character, thus can't make it available for commerical use.
描述:
训练词语: ChuXL
名称: ChuckieXL_PonyXLV6_v1.safetensors
大小 (KB): 223100
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success