Introducing Bland Speech v3, the most realistic voice model.

Introducing Bland Speech v3

The most human
text-to-speech model.

#1 model on Audio Realism Bench.
Speaker 1
170/1000
0:00 / 0:00
Start freeRead the docs

Blind listeners agree.

Second only to real humans.

Audio Realism Bench

Elo rating
  1. 1,501Human speech
  2. 1,384Bland Speech v3
  3. 1,237MAI-Voice-2
  4. 1,177Grok TTS
  5. 1,154GPT Realtime 2
  6. 1,126Gemini 2.5 Pro TTS
  7. 1,105Speech-02 HD
  8. 1,098Sonic 3.5
  9. 1,048Gemini 3.1 Flash TTS
  10. 1,011GPT-4o Mini TTS
  11. 1,003Gemini 2.5 Flash TTS
  12. 968Lightning v3.1 Pro
  13. 956Eleven v3
  14. 785TTS-1.5 Max

Audio Realism Bench is intelligence.ai's blind listening benchmark. Each trial plays two models reading the same script, unlabelled, and vetted native listeners pick which sounds like a real person. Results compound into Elo ratings, with real human recordings hidden in the pool as the control. Bars start at 600, not zero.

Hear for yourself.

KarenCurated · V3

Thanks for calling Home Appliance Solutions. Um, how can I help you today?

Can you uh repeat the, can you repeat your phone number?

You're welcome, have a good one, bye now.

Create your own voice.

Unbelievably human clones from just 10 seconds of audio. You confirm every voice is yours to use.

Ship it.

Your agents have never sounded this human.

curl -X POST "https://api.bland.ai/v1/speak" \
  -H "Authorization: Bearer $BLAND_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Thanks for calling. How can I help you today?",
    "voice_id": "VOICE_ID",
    "output_format": "pcm_44100"
  }' --output speech.wav
Model
Bland Speech v3
Endpoint
POST /v1/speak
Streaming
HTTP chunked · WebSocket
Output
PCM16 WAV · 44.1 kHz
Auth
Bearer API key

Or point a coding agent at the docs.

Read the docs
Limited-time launch offer

One rate.

$0.015

per 1,000 characters. Pay as you go.

60-second note$0.015
10-minute narration$0.135
1 million characters$15

Free to start

133k characters free. 2+ hours of speech. Same rate in the studio and the API.

The $5 unlock

Load $5 once to unlock professional voice cloning and 5x concurrency.

Start freeRead the docs

FAQ

Bland Speech turns text into speech that sounds like a person. You get a studio to try it in and one endpoint to build on, /v1/speak. Type the words yourself or let the director write them.

It was built for phone calls, where polish reads as fake. It keeps the breaths, stumbles and pauses real people make. In blind listening tests on Audio Realism Bench, only recordings of real humans ranked higher.

Instructions in brackets, like [laughs] or [clears throat]. The model performs them instead of reading them. Type [ in the studio and it completes the tag for you.

Yes. 10 seconds of audio gets an instant clone. 30 minutes of verified audio gets a professional clone. Cloning unlocks once you have loaded $5 of credit, and you confirm every voice is yours to use.

One rate: $0.015 per 1,000 characters, directed or not, studio or API. New accounts start with 133k characters free, about 2 hours of speech. Credits do not expire.

POST your text to /v1/speak with an API key. Audio streams back over HTTP chunks or a WebSocket. The whole reference lives at docs.bland.ai, or point a coding agent at it.

Judge it yourself.

Start freeRead the docs