This LoRA model was trained to convert various characters into the Bluey art style, designed to maintain compatibility with other character-specific LoRA models. During the labeling process, we intentionally avoided using Bluey character names while emphasizing descriptors related to art style, settings, and detailed visual elements.
The training utilized 999 watermark-free images sourced directly from Bluey episodes. All image tagging/annotation was processed using Claude 3.5 Sonnet, with particular focus on:
1. Preserving the show's signature thick outlines and colored pencil textures
2. Maintaining pastel color palettes and soft lighting effects
3. Capturing the distinctive squat proportions and exaggerated expressions
4. Recreating authentic Australian suburban environments
5. Retaining playful domestic chaos in scene compositions
Before you downvote this LoRA:
This LoRA is trained with Illustrious v1.0 as the base model, not v0.1, meaning that you cannot run it locally, unless you paid $10 and had Illustrious v1.0 on your device. This LoRA is not compatible with SDXL and PonyDiffusion. For now, most people have to use online service to use this model.
However, on-site generation on Civitai seems to be broken, I failed to generate good pictures here. No matter how I tweak the hyper parameters, the outputs are just bad. However, on TensorArt or TusiArt, it works normally, even when the hyper parameters remain the same as on Civitai. The example pictures are all generated there.
So please try this LoRA there, or run it locally if you have Illustrious v1.0 on your device.