The AINA program has recently presented the first speech synthesis model that offers text-to-speech interpretation in the main dialectal variants of catalan. The 'Matxa' model includes central catalan, north-western, valencian and balearic and has been developed at the Unitat de Tecnologies del Llenguatge del Barcelona Supercomputing Center - Centro Nacional de Supercomputación. Currently, work is being done to incorporate aranese. The particularity of the model is that it maintains the naturalness and characteristics of each dialect because the voices chosen have been designed to ensure their quality. In total, the 'Matxa' model incorporates eight speakers with different characteristics, two voices per dialect. The model is available for testing at Hugging Face, a community of people creating artificial intelligence tools that use open source. In essence, AINA seeks to promote the use of catalan in the digital world and encourage citizen participation.
The 'Matxa' model includes balearic, valencian, north-western and central catalan and has been developed in the Unitat de Tecnologies del Llenguatge del Barcelona Supercomputing Center.