VoxForge collects high fidelity spoken audio for use with Free and Open Source, Speech Recognition Engines in many languages.
The audio is recorded at the highest quality sampling rate available to your browser/operating system. It is recorded in WAV format, an uncompressed, lossless, audio format. Most open corpora are recorded in lossy formats (such as mp3 or vorbis).
The audio collected is not necessarily 'clean speech' but is meant to reflect audio conditions 'in the wild' in order to create better acoustic models, and to provide a better test set for evaluating Speech Recognition Engines.