This dataset contains extracted and cleaned Midjourney text prompts from Discord message JSON files.
Extracted prompts from messages with double-star formatting (
prompt)
Removed embedded image URLs (e.g.,
https://s.mj.run/xxx)
Removed duplicate and invalid prompts
Split into train/validation/test sets
train.jsonl: 53,610 prompts (80%)
validation.jsonl: 6,701 prompts (10%)
test.jsonl: 6,702… See the full description on the dataset page:
https://huggingface.co/datasets/suriyagunasekar/midjourney-text-prompts.