Successful cases
New TTS system for Naiz Irratia:
We developed a new text-to-speech (TTS) system in Basque for radio broadcaster Naiz Irratia. It allows users to listen to any digital text in podcast format. The system aims to offer a catalogue of custom voices to improve the expressiveness of the voice-overs.
New TTS system for Naiz Irratia:
Overview
We developed a new text-to-speech (TTS) system in Basque for radio broadcaster Naiz Irratia. It allows users to listen to any digital text in podcast format. The system aims to offer a catalogue of custom voices to improve the expressiveness of the voice-overs.
Challenge
The main challenge was to achieve natural and expressive text-to-speech, adapted to the complex characteristics of journalistic texts (long sentences, numerous entities), requiring few recordings per speaker, and to ensure that Naiz Irratia could select their own catalogue of voices.
Collaboration
It was developed for Naiz Irratia.
Result
A total of 150 sentences per speaker were compiled and a multi-speaker model, FastPitch + HiFi-GAN, was trained on these recordings. To improve expressiveness, versions incorporating POS tagging and BERT vectors were also trained and evaluated. However, as the original FastPitch model obtained the highest quality results, this was the one that went into production.
