Test of 200 Hand-Drawn Style LoRAs Finds 25% Ineffective
diffusion_throwaway · reddit · 2026-08-27
A user tested 200 LoRAs designed to mimic traditional hand-drawn art styles using a fixed prompt and parameters. The results suggest about 25% of the LoRAs failed to apply the intended artistic style effectively at standard strength, raising questions about the need for prompt engineering or potential cherry-picking by creators.
More from Multimodal
- Foundation Stereo pointcloud achieves near-imperceptible latency — yacineMTB · 2026-08-27
- Demo: Setting the Mood of Danger with Sound Effects and Music — techtech248 · 2026-08-27
- Nvidia reportedly buying Hugging Face; users scramble to hoard local models — MedicalPhase6556 · 2026-08-27
- Mythos animates a crying face on 3D Gaussians after calling it the hardest effect — repligate · 2026-08-27
- Cinematic architectural photography generated with AI — aziz4ai · 2026-08-27
- Manycore launches Lux3D, a 3D generative model with 20s generation — _akhaliq · 2026-08-27