Text-to-Speech-using-Tacotron2

Converting text to audio and applying audio augmentation

Topic

In this notebook I will experiment with speech synthesis from text using Tacotron2 which is a deep neural network that uses sequence to sequence architecture and which produces a mel spectromgram out of a text and then converts it to audio using Vocoder. After extracting the audio I will then use a series of audio augmentation techniques to make the sound more natural. So let's get started !

Objectives

• Convert text to audio
• Use audio augmentation to enhance the extracted audio

Summary

• Importing libraries
• Text to speech
• Adding white noise
• Time stretching
• Pitch scaling
• Inverting polarity
• Random gain
• Conclusion

Libraries

• Torchaudio
• Numpy
• Ipython
• Librosa
• Matplotlib