| --- |
| pretty_name: Midjourney Sparse Prompts |
| language: |
| - en |
| size_categories: |
| - 100K<n<1M |
| tags: |
| - text |
| - prompts |
| - midjourney |
| - image-generation |
| configs: |
| - config_name: default |
| data_files: |
| - split: train |
| path: data/train-00000-of-00001.parquet |
| license: mit |
| --- |
| |
| # Midjourney Sparse Prompts |
|
|
| 461,489 unique English-language location prompt strings, each derived from combinations of Magic: The Gathering (MTG) card concepts. |
|
|
| Kept purposely sparse, these prompts can be used as-is for exploratory image generation with Midjourney and any other image-generation model, or as grounded semantic anchors for VLMs and LLMs. |
|
|
| ## Example outputs from Midjourney |
|
|
| | | | | |
| |:---:|:---:|:---:| |
| |  |  |  | |
| | Profile: `--profile lq7svk6` | Profile: `--profile 11qmvhf` | Profile: `--profile lw7si37` | |
| |  |  |  | |
| | Profile: `--profile mrpig6q` | Profile: `--profile mrpig6q` | Profile: `--profile 11qmvhf` | |
|
|
| ## Dataset format |
|
|
| 461,489 prompts in one `train` split, stored as Zstandard-compressed Parquet. Each row contains an `id` (64-bit integer) and a `prompt` (string). IDs range from `0` to `461488`, assigned in the original prompt order; preserve these IDs when filtering or reordering. The prompts retain their original ordering and wording. |
|
|
|
|
