MONET: 104.9 Million Curated Image-Text Pairs for Text-to-Image Training

MONET is a large-scale, curated image-text dataset designed for training text-to-image systems. It contains 104.9 million high-quality image-text pairs distilled from 2.9 billion raw pairs across nine open sources. The dataset was created by jasperai and was last updated on the platform in May 2026.

Use Cases

• Training text-to-image generation models based on the described high-quality image-text pairs.
• Benchmarking model performance on a large-scale, filtered multimodal dataset.
• Conducting research on dataset curation methods like safety filtering and duplicate removal mentioned in the description.
• Fine-tuning diffusion models or other generative architectures using enriched captions.

Strengths

• Contains 104.9 million high-quality image-text pairs.
• Curated from a massive pool of 2.9 billion raw pairs.
• Underwent successive stages of safety filtering, domain-based filtering, and duplicate removal.
• Includes re-captioning with multiple sources to enrich the text data.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect source bias inherent to the nine heterogeneous open sources used.

Provenance

Source

Distilled from nine heterogeneous open sources (6 real and 3 synthetic).

Collection Method

Successive stages of safety filtering, domain-based filtering, exact and near-duplicate removal, and re-captioning.

Time Range

null

Freshness

Last updated 2026-05-07 11:54:42; freshness should be verified.

Geography

null