Civitai Stable Diffusion 337K: 337,000 Image-Prompt Pairs

337,000 image-text pairs sourced from the Civitai platform by thefcraft, last updated in December 2024. The collection links Stable Diffusion generated images with their corresponding generation metadata via a shared identifier.

Use Cases

• Fine-tuning text-to-image models using the prompt text and corresponding images
• Analyzing prompt engineering patterns across 337,000 records
• Training reward models by evaluating image quality against generation parameters linked by the 'id' field

Strengths

• 337,000 record scale
• Direct mapping via 'id' field
• Parquet format for metadata efficiency

Limitations

• Requires manual extraction and joining of zip files
• No explicit safety filtering or quality labels provided
• Unknown license status

Provenance

Source

Civitai

Collection Method

Scraped

Freshness

Last updated December 31, 2024.