Speech-to-Text

When I discuss the available sources of information with customers, since breaking information silos is one of my priorities at Google, they often leave audio and speech data out of the picture, even if they usually have digital call centers with hours of support conversations recorded.

Adding speech as an additional source of information can help us access more data and obtain better insights by knowing our customers better. Besides, text is much easier to analyze than audio.

Speech-to-Text can help us transcribe audio to text and offers some useful features, such as specific enhanced models for phone calls and videos, multiple speaker labeling and splitting, automatic language detection, or word-level confidence scoring. Multiple audio formats are supported, such as WAV, MP3, FLAC, AMR, OGG, or WEBM.

Since the length of a sound file can vary, transcriptions can be performed synchronously or asynchronously. In the first case, our code will wait for the process...