Text to Speech In Unity

Important! The current model (fastspeech) does not work well with short phrases. (e.g. "hi", "how are you", etc.)

This package provides a fully functional cross platform Text To Speech engine using deep learning models integrated in Unity with C#!

You can find the example repository here.

• Text to Speech In Unity
  • Text To Speech
  • Installation
  • Support the project!
  • Join the community!
  • License
  • References

Text To Speech

The model that we use for TTS is FastSpeech. The TFLite model that we used is converted from a pre-trained model found in the TensorflowTTS repository.

To prevent Unity from freezing when inferencing the TFLite model, we run the inference process in a new thread and play the audio in the main thread once it is ready.

Installation

External dependencies:

• voxell.util (UnityUtil)
• com.github.asus4.tflite (UnityTFLite)

1. Clone the UnityUtil repository into your Packages folder.
2. Add TFLite package via "add package from git URL" selection in the package manager using this link: https://github.com/asus4/tf-lite-unity-sample.git?path=/Packages/com.github.asus4.tflite
3. Clone this repository into your Packages folder.
4. Download the TFLite models from Google Drive and import them into Unity (place them inside the Assets/StreamingAssets folder).
5. And you are ready to go!

Support the project!

Join the community!

License

This repository as a whole is licensed under the GNU Public License, Version 3. Individual files may have a different, but compatible license.

See license file for details.

References

1. Fastspeech
2. MelGAN
3. All TFLite model inferencing will not be possible without the help of the Unity TFLite repository.