StyleTTS2 Catalan Multispeaker
BSC's Catalan StyleTTS2 release trains a multispeaker speech synthesiser for Catalan.
Model and training
The card describes training from scratch with the StyleTTS2 method, using 14.8 hours for training and 0.2 hours for validation at 24 kHz. It uses Catalan alignment and prosody components, alongside the original pitch extractor and Microsoft WavLM. The listed LibriTTS and Kokoro fine-tunes are comparison experiments, not the ancestry of the main checkpoint.StyleTTS2 Catalan Multispeaker: official model card
Practical use
Practical guidance. Use Catalan text and inspect pronunciation, rhythm and speaker consistency. Short training clips may not establish reliable long-form narration. Keep comparisons with the card's experimental fine-tunes clearly labelled, and check the GPL license before distributing a model-based application.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | styletts2-catalan-multispeaker |
| License | GNU General Public License 3.0.StyleTTS2 Catalan Multispeaker: official model card |
Related models
Matxa-TTS v2 Central Catalan is BSC's grapheme-based speech model built with the Matcha-TTS flow-matching approach.
Launch date not confirmedDownloadable- StyleTTS2 Spanish Multispeaker
Barcelona Supercomputing Center · BSC StyleTTS2
BSC's Spanish StyleTTS2 model generates expressive speech from Spanish text.
Launch date not confirmedDownloadable