Skip to content

Dataset · vivym

Cleaned Midjourney Prompts Across Versions

vivym/midjourney-prompts

Over 9 million cleaned text prompts for the Midjourney image generator versions 1.0 through 5.2.

What's inside

The dataset contains 9,085,397 cleaned English prompts categorized by Midjourney version (from 1.0 to 5.2, with the most for 5.0 and 5.2) and style (default, raw, expressive, scenic, cute, original). In addition to text and tabular metadata, image modality is specified in the entry.

What you can do with it

  • Training language models for automated prompt enhancement and expansion
  • Training and evaluating text-to-image models and LoRA adapters
  • Analyzing the evolution of user prompts across different Midjourney versions
Tasks
Text → image
Languages
English
Size
1M–10M rows
Downloads this month
265
Likes
54

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("vivym/midjourney-prompts")