Victoria: the first synthetic voice for sports news in Spanish

“Victoria, la voz del fútbol” (“Victoria, the voice of soccer”) is a pioneering project developed by Monoceros Labs for PRISA Media (Cadena SER and AS) and a milestone in synthetic content creation in Spanish. Victoria is a custom synthetic brand voice whose identity was generated by artificial intelligence. She reads sports news and lets users interact through Alexa to get up-to-date information about their favorite teams.

The challenge

The project posed several technical and creative challenges:

• Designing a unique synthetic voice that wouldn’t identify any real person.
• Developing different prosody styles for different contexts (conversation and news reading).
• Maintaining the voice’s quality and naturalness across different use cases.
• Integrating the technology with multiple platforms (Amazon Alexa, web, radio).

Our solution

Designing and developing the synthetic voice

Creating Victoria’s voice involved an innovative, meticulous methodology:

1. Defining the personality: we worked with PRISA’s teams to define the personality traits and characteristics of the voice they wanted.

2. Recording and data preparation:

   • More than 12 hours of studio recording.
   • More than 4,200 specially designed sentences.
   • 4 hours of clean audio after processing.
   • Weeks of phonetic review and data preparation.

3. Training an advanced speech synthesis (TTS) model:

   • We used our own technology, including generative methods such as GANs.
   • The model went through more than 350,000 training iterations.
   • It took 6 days of training to obtain the final model.
   • We experimented with different configurations of voice traits and prosody styles.

4. Customization and optimization:

   • Two different prosody styles (conversational and news).
   • A phonetic dictionary with more than 3,000 sports terms, including names of players, stadiums and cities.
   • Specific adaptation for using the voice on Alexa.

5. Using the voice:

   • Through our online text-to-speech editor, Fonos.
   • Through an API, customized for the different uses and channels.

Technical features

Victoria’s voice is a custom voice with specific characteristics that make it ideal for its use case:

• Energy and pace tuned to hold the listener’s attention.
• A medium-low pitch that conveys credibility and trust.
• Prosody adapted specifically for sports news.
• A natural conversational style for interaction on Alexa.
• Audio quality adapted to multiple distribution channels.

Results and impact

The project, launched in November 2022, achieved significant results:

• More than 100,000 user interactions on Alexa.
• About 100 hours of audio generated per month.
• Between 2,000 and 5,000 news stories a month from AS.com, processed in real time.

The project’s innovative nature brought several milestones and recognitions:

• The first AI-generated synthetic voice in Spanish created with this methodology.
• Sponsorship by a leading car brand.
• The “Best New Audio and Voice Product 2023” award at INMA’s Global Media Awards.

Technological innovation

The project stands out for several innovative aspects:

• Our own technology: a unique architecture for speech synthesis in Spanish, with generative AI models, funded by CDTI through NEOTEC.
• Multimodality: able to work in different contexts (radio, virtual assistant, web).
• Scale: a system able to process large volumes of text news and generate audio news in real time.
• Customization: precise adaptation of the voice to different styles and contexts.

Impact on the industry

This project put PRISA and Cadena SER at the forefront of synthetic content creation (synthetic media) and opened up new possibilities for reaching new audiences. It’s a success story in applying AI to create innovative, accessible user experiences.

More information

• Our blog post about Victoria
• Victoria, la voz del fútbol (in Spanish)
• Victoria, Cadena SER’s synthetic voice, best international audio product of the year (El País, in Spanish)
• “Victoria, la voz del fútbol,” Carrusel Deportivo’s new virtual contributor (AS, in Spanish)

This project was funded by CDTI as part of the project “Speech synthesis in Spanish for building natural conversational systems.”