# High-Fidelity Speech Synthesis

A 2.2 billion parameter text-to-speech model trained on 45,000 hours of audio data, supporting voice feature control via text prompts

Vocos is a fast neural vocoder that achieves efficient audio reconstruction by generating spectral coefficients, particularly suitable for text-to-speech tasks.

This is a Japanese text-to-speech (TTS) model trained on the ESPnet2 framework, using the VITS architecture, completed by mio on the amadeus dataset.

A FastSpeech2 text-to-speech model trained on the ESPnet framework and talromur dataset, supporting Icelandic speech synthesis.