Skip to main content
POST
/
v1
/
speak
Speak
curl --request POST \
  --url https://api.bland.ai/v1/speak \
  --header 'Content-Type: application/json' \
  --header 'authorization: <authorization>' \
  --data '
{
  "text": "<string>",
  "voice_id": "<string>",
  "output_format": "<string>",
  "stream": true,
  "language": "<string>",
  "consistency": 123,
  "expressiveness": 123,
  "boost": 123
}
'
HTTP/1.1 200 OK
Content-Type: audio/x-wav
Content-Length: 98806
x-latency: 396
x-cost: 0.001

<WAV binary>

​
Overview

Synthesizes text into a WAV audio file using a Bland TTS voice. Set stream: true to receive the audio as a chunked response while it’s being generated.
Requires a Bland TTS voice (BTTS, BTTS_V2, or BTTS_V3).
Every generation is automatically stored and retrievable via List TTS Generations and Get TTS Generation.
For lower-latency streaming with chunked WAV-header backfill, use Stream Speech instead.

​
Pricing

Text-to-speech pricing scales with plan:
PlanRate
Start$0.02 per 100 characters
Build$0.02 per 150 characters
Scale$0.02 per 200 characters
Enterprise$0.02 per 400 characters
Some voices in the public library carry an additional per-character creator fee. The exact cost for a generation is returned in the x-cost response header.

​
Headers

​
authorization
string
required
Your API key for authentication.

​
Body Parameters

​
text
string
required
The text to synthesize. Maximum 5,000 characters per request.Supports pause markers in the form <|N|> where N is a duration between 0.1 and 10.0 seconds, for example: "Welcome to Bland. <|0.8|> Let's get started."
​
voice_id
string
required
ID of the Bland TTS voice to use. Pass either the voice UUID from List Voices or a curated voice name (for example willow, juniper, valentine_experimental).
​
output_format
string
default:"pcm_44100"
Audio container/sample rate of the response.
  • pcm_8000, 8 kHz PCM16 mono (telephony)
  • pcm_16000, 16 kHz PCM16 mono
  • pcm_24000, 24 kHz PCM16 mono
  • pcm_44100, 44.1 kHz PCM16 mono (default, studio)
  • ulaw_8000, 8 kHz u-law mono (telephony)
​
stream
boolean
default:"false"
When true, the response is sent with Transfer-Encoding: chunked as audio is generated. The first 44 bytes are a WAV header with placeholder sizes (0xFFFFFFFF). Subsequent chunks are PCM16 audio data. The client backfills bytes 4-7 (RIFF chunk size minus 8) and 40-43 (data chunk size) after the stream closes.
​
language
string
default:"en"
Language code for synthesis. Defaults to the voice’s primary language. Available languages depend on the voice’s underlying model (V2 and V3 voices support 17+ languages).
​
consistency
number
Voice consistency control.
  • For BTTS (V1) voices, a float between 0.0 and 1.0. Higher values produce more consistent output.
  • For BTTS_V2 and BTTS_V3 voices, an integer between 1 and 32 (per_decode). Lower values produce more consistent output.
​
expressiveness
number
Expressiveness control for BTTS (V1) voices only. Float between 0.0 and 1.0. Higher values produce more expressive speech.
​
boost
integer
Expressiveness boost flag for BTTS_V2 and BTTS_V3 voices. 0 (off) or 1 (on).

​
Response

Returns a binary WAV file with Content-Type: audio/x-wav. Inspect response headers for latency and cost.
​
x-latency
string
Time in milliseconds from request to first audio byte (streaming) or full response (non-streaming).
​
x-cost
string
Cost in USD for the synthesis, matching what was billed.
​
Content-Type
string
Always audio/x-wav on success.
​
Content-Length
string
Total bytes in the audio (non-streaming only).
​
Transfer-Encoding
string
chunked when stream: true.
HTTP/1.1 200 OK
Content-Type: audio/x-wav
Content-Length: 98806
x-latency: 396
x-cost: 0.001

<WAV binary>
Docs for agents: llms.txt