Overtone · text to speech

A voice that means it.

Write the feeling into the line → [joy], [sigh], [whisper]. Overtone performs it, streamed while the sentence is still being written.

Get an API key Hear the voices

17 emotions16 sounds48 kHz streaming

[tender] I read your letter again last night. [sigh] [whisper]I kept it.[/whisper]

STREAMING
48 kHz
OPENAI-COMPATIBLE
Ten voices

Pick a voice. It answers.

Real output, unedited. Tap one; tap again to stop.

Tags

Direct the performance, inline.

An emotion at the start sets the mood of the whole line. Sounds land where you put them. Delivery styles wrap the words they change.

17
emotions, from joy and tender to grief, fear and tired
16
sounds: laughing, sigh, gasp, sniff, crying, yawn and more
5
delivery styles: whisper, low voice, shouting, tearful, trembling
Two modes

Best when it’s heard twice. Fast when it’s live.

Best renders several takes of every sentence and keeps the cleanest. Fast takes one and gets the first audio out sooner, for agents and live calls: about 150 ms to first audio at the API, and 353 ms median end to end through our edge in our October 2026 measurements.

English today, with ten American voices.

# Overtone, with the OpenAI SDK pointed at Yoob
from openai import OpenAI
client = OpenAI(base_url="https://tts-api.yoob.com/v1", api_key=YOOB_API_KEY)

with client.audio.speech.with_streaming_response.create(
    model="yoob-tts-1", voice="w1243",
    input="[joy] You made it! [laughing] I knew you would.",
) as r:
    r.stream_to_file("hello.wav")

Give your product a voice that means it.

Get an API key Read the docs Talk to sales