TEXTUAL INVERSION TESTS (~11/18/2022)

Various old TIs trained on the following characters:

Girl's Frontline:

  • Persicaria (~200 images)
  • P90 (~200 images)
  • Springfield (~150 images)
  • Negev (~100 images)
  • KAC-PDW (~11 images)
  • FMG9 (~10 images)

Doki Doki Literature Club:

  • Monika (~25 images)
  • Yuri (~25 images)
  • Sayori (~20 images)
  • Natsuki (~20 images)

All were trained on NAI model, with as little as 5000 steps up to 20000 steps.

(uploaded for archiving purposes: may look for older versions and more params/hyperparams used with these)

PREVIEWS (~08/14/2023)

Most generated with Silicon29 and a custom model based off it, using simple prompts and nothing else, following the format:

1girl, (solo:1.2), <TI>, <char booru tag>, <color of> eyes, best quality, <one or two additional prompts, like "dress">

neg prompt: (cropped:1.4), (loli:1.5), (child:1.5), text, low quality, normal quality, deformed, (bad-hands-5:1.2), (FastNegativeV2:1.1), deep fried, nsfw, big ribbon, red ribbon, painting, jpeg artifact, deformed, red ribbon, too many ribbons, messy ribbon, braids, umbrella, floating hair, see-through clothing, torn clothing, bad hair, gross, necklace, bad proportions, text, username, <one or two additional prompts, like "school uniform">

Girl's Frontline (best to worst):

Doki Doki Literature Club (best to worst):

It's ~2022's files in near late 2023 SD world (as of writing). Maybe it's obsolete, but might be of mild interest.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.
The model cannot be deployed to the HF Inference API: The model has no library tag.