Text to speech in nine languages.

One sentence, nine languages.

• Sophie · Spanish

  Hola, te llamo de la clínica para confirmar tu cita de mañana a las diez y media. ¿Te sigue viniendo bien, o prefieres otra hora?

• Maria · Catalan

  Hola, et truco de la clínica per confirmar la teva cita de demà a les deu i mitja. Encara et va bé, o prefereixes una altra hora?

• Sophie · Galician

  Ola, chámote da clínica para confirmar a túa cita de mañá ás dez e media. Aínda che vén ben, ou prefires outra hora?

• Maria · Basque

  Kaixo, klinikatik deitzen dizut bihar hamar eta erdietan duzun hitzordua berresteko. Oraindik ondo datorkizu, edo beste ordu bat nahiago duzu?

• Emma · English

  Hi, I'm calling from the clinic to confirm your appointment tomorrow at half past ten. Does that still work for you, or would another time be better?

• Sophie · Portuguese

  Olá, ligamos da clínica para confirmar a sua consulta de amanhã às dez e meia. Ainda é um bom horário, ou prefere outro?

• Maria · French

  Bonjour, je vous appelle de la clinique pour confirmer votre rendez-vous de demain à dix heures et demie. Cela vous convient toujours, ou préférez-vous un autre horaire ?

• Emma · Italian

  Buongiorno, la chiamo dalla clinica per confermare il suo appuntamento di domani alle dieci e mezza. Le va ancora bene, o preferisce un altro orario?

• Emma · German

  Guten Tag, ich rufe aus der Praxis an, um Ihren Termin morgen um halb elf zu bestätigen. Passt Ihnen das noch, oder hätten Sie lieber eine andere Uhrzeit?

Spanish · Catalan · Galician · Basque · English · Portuguese · French · Italian · German

A model that speaks, not a library of sounds.

Text to speech turns written text into spoken audio. Versa is a neural model that learned how people speak from recorded speech. It uses flow matching: each chunk of audio is generated in parallel rather than one token at a time, which is why it starts speaking fast and stays small: 91.4 million parameters in all.

• Any voice, any languageA voice is a timbre, not a language: all ten voices speak all nine languages.
• Your own voice3 to 30 seconds of speech becomes a new voice, with no training.
• StreamingAudio arrives while the rest is still being made, over WebSocket or the SDKs.
• Price$4.10 per million characters, prepaid, no plan.

One request.

curl -X POST "https://api.lokutor.com/tts/synthesize" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text":"Hola, ¿en qué puedo ayudarte?","voice":"F2","language":"es"}' \
  --output speech.wav

import requests

r = requests.post(
    "https://api.lokutor.com/tts/synthesize",
    headers={"Authorization": "Bearer YOUR_API_KEY"},
    json={"text": "Hola, ¿en qué puedo ayudarte?", "voice": "F2", "language": "es"},
)
open("speech.wav", "wb").write(r.content)