Round logos, icons or portraits

This model creates round "logos" and was trained on a wide range of subjects from pop culture characters, animals, places, objects, etc. Generally in a flatter color style with bold outlines.

• V2 was trained with a mix of natural language and danbooru tags on Animagine XL v3

  • Retrained on a pruned dataset of 250 images.

  • Helpful Positives: white/black/gold border, dark/blue/green/brown/gradient background, cartoon character

  • Example Prompt: round logo, SUBJECT, TAG1, TAG2, ... COLOR border with COLOR background

• V1 was trained with danbooru tags on NAI.

  • Trained on a dataset of 1500 images.

  • Helpful Positives: no humans, solo, simple background, gradient background

Use the included ControlNet images for consistent results! Download

• Set preprocessor to SoftEdge or None

v1 was used to generate the banner logo for my published article here:

R-ESRGAN-AnimeVideo-UI (reav-ui)

(ControlNet was not used for provided samples)