ArtPrompter

A gpt2 powered predictive algorithm for making descriptive text prompts for A.I. image generators (e.g. MidJourney, Stable Diffusion, ArtBot, etc). The model was trained on a custom dataset containing 666K unique prompts from MidJourney. Simply start a prompt and let the algorithm suggest ways to finish it.

from transformers import pipeline

prompter = pipeline('text-generation',model='pearsonkyle/ArtPrompter', tokenizer='gpt2')

texts = prompter('A portal to a galaxy, view with', max_length=30, num_return_sequences=5)

for i in range(5):
    print(texts[i]['generated_text']+'\n')

Intended uses & limitations

Build sick prompts and lots of them.. use it to make animations or a discord bot that can interact with MidJourney.

Examples

The entire universe is a simulation,a confessional with a smiling guy fawkes mask, symmetrical, inviting,hyper realistic
a pug disguised as a teacher. Setting is a class room
I wish I had an angel For one moment of love I wish I had your angel Your Virgin Mary undone Im in love with my desire Burning angelwings to dust
The heart of a galaxy, surrounded by stars, magnetic fields, big bang, cinestill 800T,black background, hyper detail, 8k, black

Training procedure

~30 hours of finetune on RTX3070 with 666K unique prompts

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e-05
train_batch_size: 16
eval_batch_size: 4
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 50

Framework versions

Transformers 4.26.0
Pytorch 1.13.1
Tokenizers 0.13.2