New AssemblyAI-TTS-1: Text-to-speech for voice agents. Learn more
Text-to-Speech API

Give your voice agent a voice callers stay on the line for

Human-sounding speech that streams from the first byte. Clone your brand’s voice from a 30-second clip and pay $15 per 1M characters, with no commitment to scale.

Start building free

No credit card required

Press play to hear it

Kyla

A young woman speaking American English at a fast pace.

per 1M characters, pay as you go
$15
languages: EN, ES, FR, DE, IT, PT
6
of audio to clone a voice
30 sec
commitment to scale concurrency
$0
Quickstart

Your agent’s first words, in minutes

  1. 1

    Create a free account

    Sign up and grab your API key from the dashboard. No credit card required.

  2. 2

    Pick a voice, or clone one

    Audition voices across languages and accents in the voice library, or clone your own from a 30-second clip.

  3. 3

    Stream speech into your agent

    Open a WebSocket, send text, and play audio as it streams back from the first byte.

Start building free

No credit card required

speak.py Read the docs
URL = (
    "wss://tts.us.assemblyai.com/v1/ws"
    "?voice_id=aai-kyla&language=english"
    "&sample_rate=24000&encoding=pcm_s16le"
)

async with websockets.connect(
    URL, additional_headers={"Authorization": API_KEY}
) as socket:
    await socket.send(json.dumps({
        "type": "Generate",
        "text": "Your order 4817 shipped this morning.",
    }))
    await socket.send(json.dumps({"type": "Flush"}))
Why AssemblyAI

Tuned for live calls, not narration

Natural to listen to, quick to respond, and ready for any call volume.

Voices that sound human

Trained on millions of hours of audio, including real conversations, so callers get answers instead of hanging up.

Streams from the first byte

Built for low latency, from the architecture to the inference engine, so your agent answers before callers fill the silence.

Concurrency that scales with you

Absorb inbound and outbound spikes on usage-based pricing. Your limits grow with call volume, with nothing to commit to up front.

Instant voice cloning

Clone your brand’s voice from a 30-second clip, with permission confirmed at upload, and put it live today.

A deep voice library

Find the voice that fits your agent across languages, accents, and regions, with new voices added continuously.

Six languages, right accent

Set the language and your agent speaks English, Spanish, French, German, Italian, or Portuguese from the first word.

Pricing

One rate, every voice

Priced by usage, scaled by usage. Your concurrency limit never requires negotiation.

AssemblyAI-TTS-1

$15 /1M characters

  • Billed on input characters, at one flat rate
  • Cloned voices cost the same as library voices
  • EU processing costs the same as US
  • No per-voice fees, minimums, or contracts
  • No spend commitment to unlock concurrency
Start building free

No credit card required

Use cases

Where every call is a first impression

  • Inbound support

    Answer order questions at any hour without dead air.

  • Appointment booking

    Confirm a time and read back the details clearly.

  • Outbound reminders

    Run reminder campaigns and let concurrency scale with them.

  • Order taking

    Repeat the order back and confirm the total during the rush.

Common questions