📘 I’m preparing documentation on creating models using unique methods.
🎁 Active users of my models will receive it for free.
Follow and share your creations.

Anime FM — tuned to vibrant colors and emotional lighting.

⚠️Excellent results with:
Schedule type: Beta
Sampling method: Euler a, DPM++ 2M, DDIM, Heun, DPM++ SDE
CFG Scale: 5-7, Sampling steps: 40+

PP (Positive Prompt):

bright, vibrant, glowing, volumetric composition, glare, reflections, volumetric glow, volumetric lighting, colored rays, detailed effects

NP (Negative Prompt):

frame, border, lowres, worst quality, monochrome, grayscale, watermark, bad-hand, big breasts, (boobs)

The problem of tiny distant faces (and hands), for example in full-body shots, can be solved by using upscaling.

_ _ _ _ _ _

... to achieve this vibrant, electronic color saturation, the dataset was prepared using a unique, highly unorthodox method. I maximized monitor brightness, laid the screen directly onto the scanner bed, and captured the images this way.

This obviously introduced artifacts: the monitor’s black bezel and brand logo sometimes got scanned as well, and the model partially learned them as features.

To reduce unwanted “monitor-like” artifacts, please add text, signature, frame, border to your negative prompt.

Enjoy the electronic glow — and don’t try this at home (probably).