When the voice fades away, silence becomes heavy. It hides emotions, memories, and nuances that are part of each person. Fran Vivó, a resident of Benaguasil and affected by amyotrophic lateral sclerosis (ALS), knows this silence well. The disease, without warning, robbed him of the ability to speak with his own voice. Today, thanks to artificial intelligence, he has regained it.

The VertexLit research group, led by Jordi Linares, has presented a project that exemplifies the transformative power of artificial intelligence applied to people's well-being. This project came to ValgrAI through documentary filmmaker Alex Badia and researcher Gema Piñero from the UPV institute iTEAM.

The researcher who volunteered to work on recovering Fran's voice was Jordi Linares, a member of the Joint Research Unit of the Valencian Graduate School and Research Network of Artificial Intelligence (ValgrAI) and the Valencian Research Institute for Artificial Intelligence (VRAIN) at the Universitat Politècnica de València (UPV).

The team faced a major challenge: reconstructing Fran's voice with just 20 minutes of audio in Spanish and Valencian, his native language. Using neural networks trained to work with limited corpora — especially in Valencian, where resources are scarce — the researchers analysed his vocal dynamics and incorporated emotional modulations. They created an adaptive model, all with the aim of offering a language rich in human traits and with the tones typical of the Benaguasil area, rather than a robotic voice.

The audio the team had to work with consisted of WhatsApp voice messages, but this was scarce material, as Fran had lost his ability to speak just as the audio system was being introduced in this messaging app.

The project developed for Fran Vivó transcends technology: the faithful recreation of Fran's voice, using an AI-based cloning system that has successfully recovered his timbre, prosody, and identity. His unique way of speaking replaces the robotic voice of an eye-controlled communication app with his own voice.

The family has played a fundamental role in this process, helping to construct speech and enabling the emotional charge, intonation and message intent through an editor that preserves Fran's vocal identity. In addition, the tool is editable allowing users to generate the text, provided they understand what he wants to say. The technology only provides the voice: the emotion is still his.

This project has not only allowed Fran to recover his identity, but also opens up a wide range of possibilities for ALS patients, helping them to regain who they are and maintain hope in life, as Fran himself acknowledges.

The VertexLit group, which has carried out this project in a totally altruistic manner, aims to show the humanising potential of AI, as it can restore lost abilities and dignify the lives of people living in extreme situations. This project does not seek to stand out for its technical complexity, but for its human impact.

Likewise, Jordi Linares, director of the VertexLit group, affirms that they seek to give visibility to people affected by ALS, since ALS continues to be an invisible disease in social and media discourse, despite the immense suffering it causes patients and families, so that 'this voice is not only for Fran. It is for all of them,' he assures.

As explained by Vicent Botti, director of VRAIN and general director of ValgrAI at the UPV, who opened the conference where this development was presented, "this project is neither an experiment nor a demonstration. It is an ethical promise: science and technology must serve those who need the most support, and just as Fran has spoken again, thousands of voices will also be heard".

The presentation of this pioneering project by the VertexLit research group took place within the framework of the 2nd Conference of the Valencian Institute for Research in Artificial Intelligence (VRAIN) at the Alcoy Higher Polytechnic School.

At the conference, researchers from different groups at the institute also presented advances in artificial intelligence in fields such as automatic subtitling, detection and extraction of web reviews, automated detection of microplastics, and security in interactions in virtual and hybrid environments.