An experimental Flipnote Studio LoRA
Trained off 50 frames from several different Flipnotes made on the DSi/3DS
EXPECT VERY WEIRD BIZARRE OUTPUTS. THIS IS EXPERIMENTAL!

Because this version was trained on a small 50 image dataset to experiment, it performs best with simple or low-detail prompts. It can struggle with very complex scenes and detailed prompts. A larger, refined dataset is planned for the future.
Recommended settings:
• LoRA strength: 0.9 – 1.5
• Recommended keywords: monochrome, black and white, dithering, white background

• Sampler: Euler/Euler a (Simple or Beta) or DPM++ 2M SDE Karras
• Try to avoid extremely detailed prompts for now

Because the v1 dataset was small and included some UI elements, the LoRA might generate:
• Unintended borders or “screen frame” shapes
• Speech bubble-like outlines
• Text-like marks or scribbled lines
• Extra dithering or pixel noise in shaded areas
• Difficulty with complex subjects or high-detail characters

These issues will be addressed in v2 with a larger, cleaner dataset.

Mini Changelog – v1

Added:
• 50-frame Flipnote frame based training set
• Core Flipnote aesthetic: jagged outlines, binary shading, aliased strokes
• Single trigger word (dsiflipnote)
My first LoRA. pls be nice :3