Introduction
Learn what HUMAIN Voice does and find the right starting page for your integration.
HUMAIN Voice transcribes speech and generates it. Transcribe complete recordings with batch, one short clip with fast, or live audio with realtime; label who is speaking while audio arrives with live diarization; and turn text into speech. Transcription covers Arabic, English, and code-switched speech; TTS voices speak Arabic and English. Voice Agent puts a hosted agent that talks with your users into your web or mobile app.
Get started
Quickstart
Install the SDK and transcribe your first recording.
Choose a workflow
Match batch, fast, realtime, diarization, or TTS to your audio.
Authentication
Create an API key and send it on every transport.
API reference
Look up every HTTP operation and schema.
Voice Agent
Add a hosted voice agent that talks with your users.
For AI agents
/llms.txtindexes every page;/llms-full.txtcontains all pages as Markdown.- Fetch any guide page as Markdown at
/en/md/<slug>, for example/en/md/quickstart. API reference pages use/en/api-reference/md/<slug>, for example/en/api-reference/md/batch/submit-transcription-job. - HTTP contracts are in the API reference. Socket.IO events are in the realtime, fast transcription, and TTS event references.
Rules of the road:
- Keep API keys on your server; never ship them to a browser or app (Authentication).
- Send realtime and diarization audio as raw PCM16 little-endian, 16 kHz, mono (audio formats).
- Poll batch jobs to a terminal status, and pass
save_result=trueif you may read the result twice (Batch REST). - End a stream with
closeand wait foris_final(End the stream). - Branch on the error
code, not the message text (error response shape).