KempnerInstituteAI/

flux.2-dev-synthetic-2M

	Flux2.dev Synthetic: 2.2M Text-to-Image Pairs at 512×512




	
	Dataset Summary

This dataset contains ~2.2 million large-scale synthetic image–caption pairs generated using the FLUX.2-dev diffusion model:

Model: black-forest-labs/FLUX.2-dev Caption source: Text2Image-2M Total samples: 2,282,665 (571 shards × ~4000 samples) Resolution: 512 × 512 Image format: PNG (lossless) Total shards: 571 Samples per shard: 4000 (last shard: 2665) Total size: ~865 GB

Each sample consists of:… See the full description on the dataset page: https://huggingface.co/datasets/KempnerInstituteAI/flux.2-dev-synthetic-2M.