superprompt-v1

SuperPrompt

A T5 model fine-tuned on the SuperPrompt dataset for upsampling text prompts to more detailed descriptions. Meant to be used as a pre-generation step for text-to-image models that benefit from more detailed prompts.

Usage

pip install transformers

from transformers import T5Tokenizer, T5ForConditionalGeneration

tokenizer = T5Tokenizer.from_pretrained("roborovski/superprompt-v1")
model = T5ForConditionalGeneration.from_pretrained("roborovski/superprompt-v1", device_map="auto")

input_text = "Expand the following prompt to add more detail: A storefront with 'Text to Image' written on it."
input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to("cuda")

outputs = model.generate(input_ids, max_new_tokens=77)
print(tokenizer.decode(outputs[0]))

# The neon sign above the storefront reads "NeurIPS" in bold, white letters. The storefront is surrounded by a bustling cityscape, with skyscrapers and neon signs lining the walls. The sign is surrounded by a variety of colorful goods, including a variety of fruits, vegetables, and fruits, all arranged in a neat and organized manner. The storefront is surrounded by a bustling crowd of people, all chatting and laughing as they go about their daily routines.

It's important that you use the exact task prefix: Expand the following prompt to add more detail: and limit the max tokens to 77 - which is the max prompt length for Stable Diffusion models.

Examples

Sample prompts taken from the DrawBench dataset.