Hugging Face's logo

hollowstrawberry
/
holotard

Text-to-Image
English
stable-diffusion
vtuber
hololive
stable diffusion 1.5
textual-inversion
lora
character
Model card Files Community
4
holotard / README.md
hollowstrawberry's picture
hollowstrawberry
Update README.md
da3bb39
preview code
|
raw
history blame
7.61 kB
metadata
license: creativeml-openrail-m
language:
  - en
pretty_name: Hololive vtuber LoRAs and TIs
task_categories:
  - question-answering
tags:
  - stable-diffusion
  - vtuber
  - hololive
  - stable diffusion 1.5
  - textual-inversion
  - lora
  - character
  - text-to-image
pipeline_tag: text-to-image
metrics:
  - character

This repository links to Hololive vtuber embeddings scattered around the web. It also serves as a backup in case they get removed. If you don't want a copy of your embedding to be hosted here, let me know.

These are meant to be used with Stable Diffusion 1.5, preferably through the stable-diffusion-webui running on your computer or Google Collab. After loading your main model (such as Anything V3 and its VAE) you can then use these embeddings by putting them in the ./embeddings/ folder and the ./models/LORA/ folder. Then, use the red button near Generate in your webui and refresh to see your embeddings.

Some discretion is advised as these embeddings may be capable of generating NSFW content, and their sources may contain more references to NSFW content.

Vtuber-related models

There exists a family of full SD models with the addition of vtuber training (known as hll). If the character you want does not have a source in the list below, or if you want to have wide access to vtuber generation, here are some notable ones:

You may also manually mix hll1 (old and reliable) or hll3.1 (new and more vtubers) with your favorite model via weighted sum.

Click here for a comparison

Characters and costumes

Shown in the format Character Name (Source 1) (Source 2)..., where each source contains 1 or more costumes. By default a source will contain a LoRA, but if it has {TI} next to it, it will be a Textual Inversion. They go into different folders and work differently, but generally they are the same except that LoRAs are more powerful and TIs are much more lightweight.

In the case where a source is a Discord server, you will find the original post in an Embeddings forum.

Note that all of these embeddings were tested with 7thHeavenMix and AbyssOrangeMix, neither of which have extensive vtuber data.

Hololive English

Hololive Indonesia

  • Ayunda Risu
  • Moona Hoshinova
  • Airani Iofifteen
  • Kureiji Ollie (Default, Olivia)
  • Anya Melfissa
  • Pavolia Reine
  • Vestia Zeta
  • Kaela Kovalskia
  • Kobo Kanaeru (Default)

Hololive Japan