Skip to content

Dataset · Yejy53

GPT-4o synthetic text-image pairs

Yejy53/Echo-4o-Image

Synthetic dataset with images generated by GPT-4o to improve text-to-image models.

What's inside

The dataset combines images and text descriptions in English. The card states a scale of 180,000 synthetic examples generated by the GPT-4o model (metadata indicates a range of 1k–10k). The data was created specifically to cover rare scenarios missing in real photo datasets.

What you can do with it

  • Fine-tuning and LoRA tuning of open-source text-to-image generators
  • Improving generation quality for rare or complex scenes via text prompt
  • Distilling the visual capabilities of GPT-4o into local models
Tasks
Text → image
Languages
English
Size
1K–10K rows
Downloads this month
718
Likes
36

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("Yejy53/Echo-4o-Image")