Development of a Text To Speech component that operates offline for the EBO educational robot.

EBO is an educational robot that has the ability to express emotions. In addition, it has a programming platform designed in blocks so that children can learn to program in a simple way.

This robot lacked the ability to speak. This project has given EBO a voice offline. This voice has been generated from a neural network that has been trained to get a good tone.

The code done: Get the code

EBO documentation: Documentation

Posts.

1. Introduction.
2. First advances.
3. The voice is taking shape.
4. Final results.

Commits.

Commits

Neural network used.

Tacotron

Example of the final version.

The example will be developed in the robot simulator. The code used for the example is:

Before launching the simulator, we will launch the component tacotron, from the component folder, as:

python3 src/tacotron.py etc/config

After this, we will launch the simulator, add the client and start the components and the simulation. After the simulation we will get a temporary audio like this one:

Audio

Examples of the voice.

Audio samples

For the following versions.

• Improve the synthesis to get the audios in a more fluid way.
• Add Spanish language.
• Modulate the audios to represent more emotions.