Overview:

Trained on 70 hand-scanned/glitched images, this LoRA produces imagery with a variety of scanographic glitch effects ranging from moderately subtle to lush lofi chaos. Originally intended for use with more photorealistic models but also plays well with Anime and Illustration focused Checkpoints and LoRAs. A great resource for those who need quick glitched assets to collage or throw some type over.

Recommended Settings:

LoRA Weight: 0.6+ | 0.7-0.8 sweet spot | 1 - 1.2 feeling spicy.

Models: v1-5-pruned-emaonly (trained on), ProtoGen_X3.4-pruned-fp16, realisticVisionV20_20, f222

VAE: vae-ft-mse-840000-ema-pruned

Clip skip: 1 or 2

Resolution: All standard resolutions up to 512 x 1024 seem fine, 1024 x 1024 is where quality seems to fall off.

Sampler: DPM++ 2M Karras

Steps: 25-50

CFG: 7

Example Prompt:

Model: v1-5-pruned-emaonly

Steps: 30, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: 3948837425, Size: 640x640

Posetive: (detailed:1.1), portrait photography in the style of <lora:scanno_Glitch_01:0.8>, (mall:1.1), golden hour, cinestill, portra, nasa backrooms, 90s, nostalgic, play area, indoors

Negative: n/a

A note on image quality:

The learned style relative to the training set is very accurate. Everything from how the images look printed on matte lightweight paper stock, to the small size of each image per page affecting the scanned output, the results are images that can appear to be lower resolution than they actually are (especially at full strength). In future versions I'll try to remedy this by scanning larger images individually, rather than many per sheet. In the meantime, however, this LoRA is a lean, mean, vibey lofi machine ;).