MSDWild Dataset

Our new Developer Platform and API prices are live!

The MSDWild dataset is a large, diverse dataset of real-world noisy speech used for training and evaluating speech enhancement models.

It contains field recordings from uncontrolled environments, ideal for stress-testing enhancement models.

It provides realistic, challenging data that helps models generalize beyond clean lab recordings.

Here at ai-coustics, we train and validate our models on MSDWild to ensure robust performance in authentic noisy conditions. Read more about it here.