A synthetic voice of its own for an energy company

The advertising department of a major Spanish energy company was already using artificial intelligence to personalize its communication. Now it wanted a brand voice of its own for internal use: to voice demos and radio spots so it could prototype campaigns before going into the studio with voice actors.

The challenge

The voice had to be exclusive. A stock voice with a different tone wouldn’t do: a new identity had to be created.

The starting point was a description (“a woman of about 35, with energy”) and a few reference ads. Everyone on the team imagined that voice differently.

On top of that, several agencies work with the company, and all of them had to be able to use the voice from one place. And there were plans for later: more voices and more languages.

What we did

We worked in phases, with sign-off at the end of each one.

1. Definition workshop with the advertising team. We started by understanding what a synthetic voice can and can’t do today. Then we built a voice moodboard together and defined who it would be used for and what for.
2. From words to voice. We translated what came out of the workshop into characteristics that can be measured in a voice.
3. A model of its own. We trained a custom speech synthesis model and searched it for identities that matched what had been defined.
4. Choosing the voice. We presented several options to listen to. This part is subjective, so the people who’ll work with the voice make the call.
5. Use in Fonos. The voice, exclusive to the company, is available in Fonos for the agencies it authorizes.