# emolia-balanced-5M-subset The version of the balanced 5M-clip Emolia subset shards used in the VoiceCLAP experiments. It differs from the current shards in [laion/emolia-balanced-5M-subset](https://huggingface.co/datasets/laion/emolia-balanced-5M-subset) (same shard names, different contents). The original dataset card is kept as `README_original.md`.