/v1/speech speaks with one of these voices. Each recording below was made through the API. Press play to hear it: nothing downloads until you do.
Every voice is used with its speaker’s consent. The API does not offer voice cloning, and a voice whose consent could not be confirmed was withdrawn: a request that names it gets
400 unknown_voice, never another voice in its place.
Choosing the language
A voice speaks its first language unless you say otherwise. To make Mouhamad speak Hausa:400 unsupported_language, with the languages it does speak in the message.
Languages
Fulfulde and Tamasheq are part of the Lab’s research, not of the API yet. We add a language to the API when we can measure that it works.