Installation:
pip install -r requirements.txt
Script:
python avatair.py
Config:
| Name | Description | Default |
|---|---|---|
| initial | Number of iterations that will run until the program is finished | 5 |
| scales | 3 types of scales for other use-cases
1 = acceptance, likeability, empathy, anthropomorphism, trust 2 = openness, conscientiousness, extraversion, agreeableness, neuroticism 3 = efficiency |
1 |
| pictures | The amount of pictures should be generated during one iteration (Int) | 1 |
| attention | Possibility to add several attention-checks after x iterations
(Array (e.g [1,2,3]), [-1] for no check) |
[1,3] |
| stablediffusion | Choose between Stable Diffusion ("sd") and Stable Diffusion XL ("xl") | "xl" |
| model | Use different types of SD-models | "SG161222/RealVisXL_V3.0" |
| token | HuggingFace login-token, just needed for some specific SD-models | "" |
| promptmodel | Generate prompt with defined or latent variables
("defined" or "latent") defined = e.g. abstraction, haircolor, eyecolor, ... latent= e.g. acceptance, likeability, ... |
defined |
Requirements:
requirements.txt
The pipeline works as following:
Paper:
Download the paper here.
License:
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
For non-commercial use only!