Hololive JP・EN・ID Collection
This is a LoRA trained on a large collection of datasets of Hololive virtual youtubers, focused on their 1st costumes.
It includes (almost) all active and former members of the Japan, English and Indonisia branches.
The model is intended for hobby level use in the creation of non-commercial fan works.
Recommended prompt:
All images were manually rated with quality/aesthetic tags:
masterpiece, best quality, low quality, very aesthetic, aesthetic.
Mature content is also tagged with nsfw, explicit tags respectively.
Positive prompt:
Where {name} is the hololive members name in lowercase, i.e. hoshimachi suisei
and {tags} being any other tags describing appearance, scene, etc.
{name}, {name} (1st costume), 1girl, virtual youtuber,
{tags},
score_9,score_8_up, score_7_up, score_6_up, masterpiece, best quality, very aesthetic
Negative prompt:
(worst quality, low quality, sketch:1.1), score_4, score_3, score_2, score_1, error, bad anatomy, bad hands, watermark, ugly, distorted, signature
Tagging
For most members, {name}, {name} (1st costume) should be effective as triggers, but extra tagging for their traits is usually required.
Some members are already known by the base model (Pony Diffusion v6) by their name, and some can also be prompted with a code recognised by the base model, i.e. aua for Houshou Marine.
