Docs Book a Demo Sign in
Text to speech

Polish text to speech, po polsku.

Text to speech in Polish for announcements, reminders, narration and customer messages, in a female and a male voice.

Listen

Two voices, in Polish.

  • Female

    Dzień dobry, w czym mogę dzisiaj pomóc?

  • Male

    Dzień dobry, w czym mogę dzisiaj pomóc?

Versa’s output for exactly the text shown, unedited, the same sentence in each of the two voices. The take was checked with a speech recogniser. Measured on eight Polish sentences, an independent recogniser scored a character error of 2.7%.

The voices are speakers from open speech corpora: Mozilla Common Voice 17, Polish (CC0). All credits.

How it works

Text in, Polish out.

Text to speech turns written Polish into spoken audio. Ours is Versa, a speech model we train ourselves: it generates speech one 70 ms frame at a time, so the first audio arrives in about 25 ms for a single request and the rest streams behind it.

Versa speaks Polish in a female and a male voice, or in a voice cloned from a few seconds of Polish audio. A voice agent can speak Polish, but our speech recognition does not understand it yet, so it cannot hold a two-way conversation in it.

  • VoicesA female and a male voice, both native to Polish.
  • Your voiceA new voice from 3 to 30 seconds of speech, no training.
  • Price$4.10 per million characters, prepaid, no plan.
  • Voice agentsAgents can speak Polish; they do not understand it yet.
The API

One request.

terminal
curl -X POST "https://api.lokutor.com/tts/synthesize" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text":"Dzień dobry, w czym mogę dzisiaj pomóc?","voice":"pl_f1","language":"pl"}' \
  --output speech.wav
speak.py
import requests

r = requests.post(
    "https://api.lokutor.com/tts/synthesize",
    headers={"Authorization": "Bearer YOUR_API_KEY"},
    json={"text": "Dzień dobry, w czym mogę dzisiaj pomóc?", "voice": "pl_f1", "language": "pl"},
)
open("speech.wav", "wb").write(r.content)

Returns a WAV file. For streaming, use the WebSocket API or the JavaScript and Python SDKs.

FAQ

Common questions.

How much does Polish text to speech cost?

$4.10 per million characters, the same in every language, billed per character from a prepaid balance. New accounts get $5 of credit with no card, about 1.2 million characters.

Which voices speak Polish?

Polish has a female and a male voice of its own. Any new voice you clone speaks Polish too.

Can I clone a voice in Polish?

Yes. Record or upload 3 to 30 seconds of speech in the dashboard and the new voice speaks Polish and the other 12 languages. Only clone a voice you own or have permission to use.

Can a voice agent hold a conversation in Polish?

Not a two-way one yet. A Lokutor voice agent can speak Polish, but our speech recognition does not understand it yet, so the caller cannot be answered in Polish. Agents hold full conversations in Spanish, Catalan, Galician, Basque, English, Portuguese, French, Italian and German.

Other languages

Spanish · Catalan · Galician · Basque · English · Portuguese · French · Italian · German · Dutch · Romanian · Russian

Hear your own text.

$5 of credit to start, no card. Type it in the speech studio and press play.