Custom Speech
Customize speech recognition models to your needs and available data. Overcome speech recognition barriers such as speaking style, vocabulary and background noise.
New to Speech service? Create a Speech resource
Improve speech recognition quality
Evaluate and improve speech recognition accuracy for your application.
Upload data
Create a project and upload your audio data, along with transcripts and related text.
Evaluate performance
Test and compare the results of Microsoft’s speech-to-text models, and evaluate the accuracy.
Improve accuracy
Improve overall and proper name recognition accuracy by training a custom model.
Deploy custom endpoint
Deploy your own custom endpoint to use in your application.
Creating a custom speech-to-text model
1 Create a Speech resource at portal.azure.com.
2 Prepare some test audio files.
3 Test Microsoft’s standard speech-to-text.
4 Upload training data (related text/audio/human transcripts).
5 Train custom speech-to-text model.
6 Deploy model.