Midjourney Prompts Across Seven Model Versions

9,085,397 cleaned text prompts for the Midjourney generative art model, spanning versions 1.0 through 5.2. The dataset includes style tags such as 'default', 'raw', and 'expressive' for a subset of prompts.

Use Cases

• Analyze prompt evolution and keyword frequency across Midjourney model versions (5.2, 5.1, 5.0, etc.) using the version counts.
• Study the adoption of specific artistic styles like 'raw' or 'expressive' versus the 'default' style among 9 million prompts.
• Train a text generation model to produce Midjourney-style prompts using the corpus of 9 million examples.
• Investigate the distribution of prompt usage over time by correlating version release dates with prompt counts per version.

Strengths

• Contains 9,085,397 prompts, providing a large-scale corpus for analysis.
• Offers version breakdowns, with 3,530,770 prompts for version 5.0 and 2,272,465 for version 5.2.
• Includes style categorizations for prompts, such as 177,953 tagged as 'raw' and 27,919 as 'expressive'.

Limitations

• No column schema or sample data is provided, making the internal structure and available features unknown.
• The dataset appears static, with no update frequency indicated beyond the last update in November 2023.
• Potential bias towards newer model versions, as versions 1.0, 2.0, and 3.0 collectively represent less than 0.2% of the data.

Provenance

Source

Midjourney, via Hugging Face user vivym.

Collection Method

Collected and cleaned prompts from the Midjourney platform.

Time Range

Covers prompts generated up to November 2023, spanning releases from version 1.0 to 5.2.

Freshness

Last updated 2023-11-15.

Geography

null