ElevenLabs · Product

ElevenAPI

The developer surface for the whole voice stack.

ElevenAPI is the same engine that powers ElevenCreative and ElevenAgents, exposed as a clean HTTP + WebSocket API. Streaming TTS, realtime STT (Scribe v2), voice cloning, dubbing and agents — all behind one API key.

Get an API key

Disclosure: We may earn a commission if you purchase through links on this page. This never affects our reviews.

Features

🚀

Streaming TTS

MP3, PCM and μ-law output with sub-second time-to-first-byte.

✍️

Realtime STT

Scribe v2 streams partial transcripts over WebSocket with VAD and diarization.

📼

Batch transcription

99+ languages, speaker labels and audio-event tags for offline pipelines.

🧬

Voice cloning API

Programmatic Instant and Professional cloning with per-tenant voice IDs.

🌐

Dubbing API

Translate and re-voice full videos with a single POST.

💵

Predictable pricing

Character-based billing with volume tiers — no surprise inference fees.

Use cases

Voice-enabled SaaS features (comments, summaries, narration)
Multilingual customer support automation
AI-native audio products and podcasting tools
Realtime transcription and captioning services
Custom voice-agent platforms built on ElevenAgents

Tutorials

Stream TTS from a Node backend

The minimum viable pattern for shipping low-latency AI voice inside a web app.

Create an API key with the tts:write scope.
Call POST /v1/text-to-speech/:voice/stream with output_format=mp3_44100_128 as a query param.
Pipe response.body straight into your HTTP response with Content-Type audio/mpeg.
On the client, play the response with new Audio(URL.createObjectURL(blob)).
Add a small warm-up ping on session start to hide TTS cold starts.

Realtime transcription with Scribe v2

Live captions and voice-input UIs with committed and partial transcripts.

Mint a short-lived realtime_scribe token server-side.
Open a WebSocket from the browser using the token.
Stream 100ms PCM chunks and render partial_transcript as it arrives.
Use VAD commit strategy for hands-free segmentation.
Persist committed_transcript to your DB — that's your source of truth.

FAQs

Is there a free tier for developers?+

Yes — the free plan includes API access with a monthly character allowance, enough to prototype end-to-end.

Does ElevenAPI support streaming?+

Both TTS output and STT input stream over HTTP chunked responses and WebSocket respectively.

Where does ElevenAPI host?+

Global by default, with EU, US and India residency available on higher-tier plans.

Resources