Yoob voice API
Speech that sounds alive, and listening that hears more than words. One API key from this console, one base URL, and endpoints shaped like OpenAI's, so existing SDKs work by changing the base URL.
Text to speech · Overtone
Expressive speech from text, with emotion, laughs, sighs and whispers written into the line. Streams as it renders.
Speech to text · Undertone
Accurate transcripts of calls, meetings and recordings up to an hour, or of live audio as it is spoken.
Emotion recognition · Undertone
How it was said: the feeling of each utterance with probabilities, laughs and sighs, whispers and shouts, and a profile of the voice.
Voice agents · Duotone
A live conversation you can interrupt, built from realtime speech to text, your language model and streaming speech.
Voice detection · Truetone
Is it a real person or a synthetic voice? A human-or-AI score for every clip and every sentence heard.
Base URL
Every request goes to one host, over HTTPS for requests and WSS for the realtime sockets:
https://tts-api.yoob.com/v1Authenticate with an API key from the API Key page, sent as Authorization: Bearer yoob_live_…. See Authentication.
Models
| Model | What it does | Endpoints |
|---|---|---|
yoob-tts-1 | Expressive text to speech, 48 kHz | /audio/speech, /audio/speech/stream |
yoob-asr-2 | The most accurate speech to text, files up to 60 minutes | /audio/transcriptions |
yoob-asr-1 | Live speech to text, and emotion recognition: feeling, sounds, voice profile and a human-or-AI score | /audio/transcriptions, /audio/transcriptions/stream |
OpenAI SDKs
Speech and transcription follow OpenAI's request and response shapes. The official SDKs work as they are for both: point them at the base URL above and pass your Yoob key. See Speech to text.
from openai import OpenAI
client = OpenAI(base_url="https://tts-api.yoob.com/v1", api_key="yoob_live_...")Yoob's own options, such as quality and tags, go in the same request; see each guide.