> ## Documentation Index
> Fetch the complete documentation index at: https://docs.cartesia.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Realtime Speech-to-Text (Auto)
> Realtime speech transcription with built-in turn detection
This endpoint is English only right now.
We expect to add more languages in the coming months.
## AsyncAPI
````yaml asyncapi.yml /stt/turns/websocket
id: /stt/turns/websocket
title: /stt/turns/websocket
description: |
Realtime speech transcription with built-in turn detection
This endpoint is English only right now.
We expect to add more languages in the coming months.
servers:
- id: production
protocol: wss
host: api.cartesia.ai
bindings: []
variables: []
address: /stt/turns/websocket
parameters:
- id: model
jsonSchema:
type: string
description: |
ID of the model to use for transcription, e.g. `ink-2`.
See [Models](/build-with-cartesia/stt/latest) for available models.
description: |
ID of the model to use for transcription, e.g. `ink-2`.
See [Models](/build-with-cartesia/stt/latest) for available models.
type: string
required: true
deprecated: false
- id: encoding
jsonSchema:
type: string
description: >
The encoding format of the audio data. This determines how the server
interprets the raw binary audio data you send.
Supported encodings: `pcm_s16le`, `pcm_s32le`, `pcm_f16le`, `pcm_f32le`,
`pcm_mulaw`, `pcm_alaw`.
For guidance on choosing an encoding, see [Audio
Input](/build-with-cartesia/stt/audio-input).
description: >
The encoding format of the audio data. This determines how the server
interprets the raw binary audio data you send.
Supported encodings: `pcm_s16le`, `pcm_s32le`, `pcm_f16le`, `pcm_f32le`,
`pcm_mulaw`, `pcm_alaw`.
For guidance on choosing an encoding, see [Audio
Input](/build-with-cartesia/stt/audio-input).
type: string
required: true
deprecated: false
- id: sample_rate
jsonSchema:
type: string
description: |
The sample rate of the audio in Hz.
description: |
The sample rate of the audio in Hz.
type: string
required: true
deprecated: false
- id: turn_start_threshold
jsonSchema:
type: string
description: >
Threshold above which to start the turn. Range: 0.5–0.9. Must stay above
the eager end threshold.
See [Configuring turn
detection](/use-the-api/stt/turns#configuring-turn-detection) for
details.
default: '0.8'
description: >
Threshold above which to start the turn. Range: 0.5–0.9. Must stay above
the eager end threshold.
See [Configuring turn
detection](/use-the-api/stt/turns#configuring-turn-detection) for details.
type: string
required: true
deprecated: false
- id: turn_eager_end_threshold
jsonSchema:
type: string
description: >
Threshold below which to eager end the turn. Range: 0.3–0.6. Must stay
between the end and start thresholds.
See [Configuring turn
detection](/use-the-api/stt/turns#configuring-turn-detection) for
details.
default: '0.4'
description: >
Threshold below which to eager end the turn. Range: 0.3–0.6. Must stay
between the end and start thresholds.
See [Configuring turn
detection](/use-the-api/stt/turns#configuring-turn-detection) for details.
type: string
required: true
deprecated: false
- id: turn_end_threshold
jsonSchema:
type: string
description: >
Threshold below which to end the turn. Range: 0.05–0.5. Must stay below
the eager end threshold.
See [Configuring turn
detection](/use-the-api/stt/turns#configuring-turn-detection) for
details.
default: '0.2'
description: >
Threshold below which to end the turn. Range: 0.05–0.5. Must stay below
the eager end threshold.
See [Configuring turn
detection](/use-the-api/stt/turns#configuring-turn-detection) for details.
type: string
required: true
deprecated: false
- id: turn_end_timeout_ms
jsonSchema:
type: string
description: >
Maximum amount of time in milliseconds that the model will wait after
the user stops speaking before ending the turn. Range: 640–11200.
See [Configuring turn
detection](/use-the-api/stt/turns#configuring-turn-detection) for
details.
default: '5600'
description: >
Maximum amount of time in milliseconds that the model will wait after the
user stops speaking before ending the turn. Range: 640–11200.
See [Configuring turn
detection](/use-the-api/stt/turns#configuring-turn-detection) for details.
type: string
required: true
deprecated: false
- id: cartesia_version
jsonSchema:
type: string
description: API version, e.g. `2026-03-01`
description: API version, e.g. `2026-03-01`
type: string
required: true
deprecated: false
bindings: []
operations:
- &ref_1
id: sendSTTTurnsAudio
title: Send s t t turns audio
type: receive
messages:
- &ref_3
id: sttTurnsAudioData
payload:
- type: string
format: binary
description: >
Send WebSocket binary messages containing raw audio data as
specified by the `encoding` and `sample_rate` query parameters.
Audio Requirements:
- Send audio in small chunks, e.g. 100 ms
- Audio format must match the `encoding` and `sample_rate`
parameters
x-parser-schema-id:
name: Send Audio Data
headers: []
jsonPayloadSchema:
type: string
format: binary
description: >-
Raw audio data as a binary message in the format specified by the
`encoding` parameter. Send audio in small chunks, e.g. 100 ms.
x-parser-schema-id:
title: Send Audio Data
description: >
Send WebSocket binary messages containing raw audio data as specified
by the `encoding` and `sample_rate` query parameters.
Audio Requirements:
- Send audio in small chunks, e.g. 100 ms
- Audio format must match the `encoding` and `sample_rate` parameters
example: '{}'
bindings: []
extensions:
- id: x-parser-unique-object-id
value: sttTurnsAudioData
- &ref_4
id: sttTurnsCloseCommand
payload:
- name: Close Command
description: >-
Send a JSON encoded close command as WebSocket text message to
close the session cleanly. All buffered audio will be processed by
the model into events.
type: object
properties:
- name: type
type: string
description: >-
Command type. Send this as a JSON encoded WebSocket text
message to close the session.
enumValues:
- close
required: true
headers: []
jsonPayloadSchema:
type: object
required:
- type
properties:
type:
type: string
enum:
- close
description: >-
Command type. Send this as a JSON encoded WebSocket text message
to close the session.
x-parser-schema-id:
x-parser-schema-id: STTTurnsCloseCommand
title: Close Command
description: >-
Send a JSON encoded close command as WebSocket text message to close
the session cleanly. All buffered audio will be processed by the model
into events.
example: |-
{
"type": "close"
}
bindings: []
extensions:
- id: x-parser-unique-object-id
value: sttTurnsCloseCommand
- &ref_5
id: sttTurnsConfigCommand
payload:
- name: Config Command
description: >-
Send a JSON encoded config command as a WebSocket text message to
update model settings.
type: object
properties:
- name: type
type: string
description: >-
Command type. Send this as a JSON encoded WebSocket text
message to update model settings.
enumValues:
- config
required: true
- name: turn
type: object
description: Turn detection settings.
required: false
properties:
- name: start_threshold
type: number
description: >
Threshold above which to start the turn. Range: 0.5–0.9.
Must stay above the eager end threshold.
See [Configuring turn
detection](/use-the-api/stt/turns#configuring-turn-detection)
for details.
required: false
- name: eager_end_threshold
type: number
description: >
Threshold below which to eager end the turn. Range:
0.3–0.6. Must stay between the end and start thresholds.
See [Configuring turn
detection](/use-the-api/stt/turns#configuring-turn-detection)
for details.
required: false
- name: end_threshold
type: number
description: >
Threshold below which to end the turn. Range: 0.05–0.5.
Must stay below the eager end threshold.
See [Configuring turn
detection](/use-the-api/stt/turns#configuring-turn-detection)
for details.
required: false
- name: end_timeout_ms
type: number
description: >
Maximum amount of time in milliseconds that the model will
wait after the user stops speaking before ending the turn.
Range: 640–11200.
See [Configuring turn
detection](/use-the-api/stt/turns#configuring-turn-detection)
for details.
required: false
headers: []
jsonPayloadSchema:
type: object
required:
- type
properties:
type:
type: string
enum:
- config
description: >-
Command type. Send this as a JSON encoded WebSocket text message
to update model settings.
x-parser-schema-id:
turn:
type: object
description: Turn detection settings.
properties:
start_threshold:
type: number
default: 0.8
description: >
Threshold above which to start the turn. Range: 0.5–0.9.
Must stay above the eager end threshold.
See [Configuring turn
detection](/use-the-api/stt/turns#configuring-turn-detection)
for details.
x-parser-schema-id:
eager_end_threshold:
type: number
default: 0.4
description: >
Threshold below which to eager end the turn. Range: 0.3–0.6.
Must stay between the end and start thresholds.
See [Configuring turn
detection](/use-the-api/stt/turns#configuring-turn-detection)
for details.
x-parser-schema-id:
end_threshold:
type: number
default: 0.2
description: >
Threshold below which to end the turn. Range: 0.05–0.5. Must
stay below the eager end threshold.
See [Configuring turn
detection](/use-the-api/stt/turns#configuring-turn-detection)
for details.
x-parser-schema-id:
end_timeout_ms:
type: number
default: 5600
description: >
Maximum amount of time in milliseconds that the model will
wait after the user stops speaking before ending the turn.
Range: 640–11200.
See [Configuring turn
detection](/use-the-api/stt/turns#configuring-turn-detection)
for details.
x-parser-schema-id:
x-parser-schema-id:
x-parser-schema-id: STTTurnsConfigCommand
title: Config Command
description: >-
Send a JSON encoded config command as a WebSocket text message to
update model settings.
example: |-
{
"type": "config",
"turn": {
"start_threshold": 0.8,
"eager_end_threshold": 0.4,
"end_threshold": 0.2,
"end_timeout_ms": 5600
}
}
bindings: []
extensions:
- id: x-parser-unique-object-id
value: sttTurnsConfigCommand
bindings: []
extensions: &ref_0
- id: x-parser-unique-object-id
value: /stt/turns/websocket
- &ref_2
id: receiveSTTTurnsEvents
title: Receive s t t turns events
description: >-
The server sends turn events as the model transcribes. Messages can be of
type `turn.start`, `turn.update`, `turn.eager_end`, `turn.resume`,
`turn.end`, or `error`. All emitted text is final — the model does not
revise previous output. The `transcript` field is cumulative within a
turn.
type: send
messages:
- &ref_6
id: sttTurnsConnected
payload:
- name: Connected
description: |
Fires once when the WebSocket connection is established.
You do not need to wait for this event before sending audio.
type: object
properties:
- name: type
type: string
description: Event type identifier.
enumValues:
- connected
required: true
- name: request_id
type: string
description: Unique identifier for this connection.
required: true
headers: []
jsonPayloadSchema:
type: object
required:
- type
- request_id
properties:
type:
type: string
enum:
- connected
description: Event type identifier.
x-parser-schema-id:
request_id:
type: string
description: Unique identifier for this connection.
x-parser-schema-id:
x-parser-schema-id: STTTurnsConnected
title: Connected
description: |
Fires once when the WebSocket connection is established.
You do not need to wait for this event before sending audio.
example: |-
{
"type": "connected",
"request_id": "2ff8af53-4d38-479d-8287-58940f01c701"
}
bindings: []
extensions:
- id: x-parser-unique-object-id
value: sttTurnsConnected
- &ref_7
id: sttTurnsTurnStart
payload:
- name: Turn Start
description: >
Marks the start of a user turn. Fires quickly after the user
begins speaking.
This event can be used to interrupt your agent to avoid talking
over the user.
type: object
properties:
- name: type
type: string
description: Event type identifier.
enumValues:
- turn.start
required: true
- name: request_id
type: string
description: >-
Unique identifier for this connection. Does not change between
turns.
required: true
headers: []
jsonPayloadSchema:
type: object
required:
- type
- request_id
properties:
type:
type: string
enum:
- turn.start
description: Event type identifier.
x-parser-schema-id:
request_id:
type: string
description: >-
Unique identifier for this connection. Does not change between
turns.
x-parser-schema-id:
x-parser-schema-id: STTTurnsTurnStart
title: Turn Start
description: >
Marks the start of a user turn. Fires quickly after the user begins
speaking.
This event can be used to interrupt your agent to avoid talking over
the user.
example: |-
{
"type": "turn.start",
"request_id": "2ff8af53-4d38-479d-8287-58940f01c701"
}
bindings: []
extensions:
- id: x-parser-unique-object-id
value: sttTurnsTurnStart
- &ref_8
id: sttTurnsTurnUpdate
payload:
- name: Turn Update
description: |
Fires repeatedly as the model transcribes the current user turn.
type: object
properties:
- name: type
type: string
description: Event type identifier.
enumValues:
- turn.update
required: true
- name: transcript
type: string
description: >-
Cumulative text for the current turn, i.e. the full text
transcribed so far in this turn, not a delta.
required: true
- name: request_id
type: string
description: >-
Unique identifier for this connection. Does not change between
turns.
required: true
headers: []
jsonPayloadSchema:
type: object
required:
- type
- transcript
- request_id
properties:
type:
type: string
enum:
- turn.update
description: Event type identifier.
x-parser-schema-id:
transcript:
type: string
description: >-
Cumulative text for the current turn, i.e. the full text
transcribed so far in this turn, not a delta.
x-parser-schema-id:
request_id:
type: string
description: >-
Unique identifier for this connection. Does not change between
turns.
x-parser-schema-id:
x-parser-schema-id: STTTurnsTurnUpdate
title: Turn Update
description: |
Fires repeatedly as the model transcribes the current user turn.
example: |-
{
"type": "turn.update",
"transcript": "Hey can you help",
"request_id": "2ff8af53-4d38-479d-8287-58940f01c701"
}
bindings: []
extensions:
- id: x-parser-unique-object-id
value: sttTurnsTurnUpdate
- &ref_9
id: sttTurnsTurnEagerEnd
payload:
- name: Turn Eager End
description: >
Fires when the model predicts that the user might be done
speaking.
type: object
properties:
- name: type
type: string
description: Event type identifier.
enumValues:
- turn.eager_end
required: true
- name: transcript
type: string
description: Cumulative text for the current turn.
required: true
- name: request_id
type: string
description: >-
Unique identifier for this connection. Does not change between
turns.
required: true
headers: []
jsonPayloadSchema:
type: object
required:
- type
- transcript
- request_id
properties:
type:
type: string
enum:
- turn.eager_end
description: Event type identifier.
x-parser-schema-id:
transcript:
type: string
description: Cumulative text for the current turn.
x-parser-schema-id:
request_id:
type: string
description: >-
Unique identifier for this connection. Does not change between
turns.
x-parser-schema-id:
x-parser-schema-id: STTTurnsTurnEagerEnd
title: Turn Eager End
description: |
Fires when the model predicts that the user might be done speaking.
example: |-
{
"type": "turn.eager_end",
"transcript": "Hey can you help me with something?",
"request_id": "2ff8af53-4d38-479d-8287-58940f01c701"
}
bindings: []
extensions:
- id: x-parser-unique-object-id
value: sttTurnsTurnEagerEnd
- &ref_10
id: sttTurnsTurnResume
payload:
- name: Turn Resume
description: >
Fires after `turn.eager_end` if the user turn has not actually
ended.
type: object
properties:
- name: type
type: string
description: Event type identifier.
enumValues:
- turn.resume
required: true
- name: request_id
type: string
description: >-
Unique identifier for this connection. Does not change between
turns.
required: true
headers: []
jsonPayloadSchema:
type: object
required:
- type
- request_id
properties:
type:
type: string
enum:
- turn.resume
description: Event type identifier.
x-parser-schema-id:
request_id:
type: string
description: >-
Unique identifier for this connection. Does not change between
turns.
x-parser-schema-id:
x-parser-schema-id: STTTurnsTurnResume
title: Turn Resume
description: |
Fires after `turn.eager_end` if the user turn has not actually ended.
example: |-
{
"type": "turn.resume",
"request_id": "2ff8af53-4d38-479d-8287-58940f01c701"
}
bindings: []
extensions:
- id: x-parser-unique-object-id
value: sttTurnsTurnResume
- &ref_11
id: sttTurnsTurnEnd
payload:
- name: Turn End
description: |
Marks the end of a user turn.
type: object
properties:
- name: type
type: string
description: Event type identifier.
enumValues:
- turn.end
required: true
- name: transcript
type: string
description: Definitive transcript for the completed turn.
required: true
- name: request_id
type: string
description: >-
Unique identifier for this connection. Does not change between
turns.
required: true
headers: []
jsonPayloadSchema:
type: object
required:
- type
- transcript
- request_id
properties:
type:
type: string
enum:
- turn.end
description: Event type identifier.
x-parser-schema-id:
transcript:
type: string
description: Definitive transcript for the completed turn.
x-parser-schema-id:
request_id:
type: string
description: >-
Unique identifier for this connection. Does not change between
turns.
x-parser-schema-id:
x-parser-schema-id: STTTurnsTurnEnd
title: Turn End
description: |
Marks the end of a user turn.
example: |-
{
"type": "turn.end",
"transcript": "Hey can you help me with something?",
"request_id": "2ff8af53-4d38-479d-8287-58940f01c701"
}
bindings: []
extensions:
- id: x-parser-unique-object-id
value: sttTurnsTurnEnd
- &ref_12
id: sttErrorResponse
payload:
- name: Error Response
description: Error information for STT WebSocket connections.
type: object
properties:
- name: type
type: string
description: Event type identifier.
enumValues:
- error
required: true
- name: error_code
type: string
description: Machine-readable error code.
required: false
- name: status_code
type: number
description: An HTTP response status code.
required: true
- name: title
type: string
description: Human-readable error title.
required: true
- name: message
type: string
description: Human-readable error message.
required: true
- name: doc_url
type: string
description: URL to relevant documentation
required: false
- name: request_id
type: string
description: Unique identifier for this websocket connection
required: false
headers: []
jsonPayloadSchema:
type: object
required:
- type
- status_code
- title
- message
properties:
type:
type: string
enum:
- error
description: Event type identifier.
x-parser-schema-id:
error_code:
type: string
description: Machine-readable error code.
x-parser-schema-id:
status_code:
type: number
format: integer
description: An HTTP response status code.
x-parser-schema-id:
title:
type: string
description: Human-readable error title.
x-parser-schema-id:
message:
type: string
description: Human-readable error message.
x-parser-schema-id:
doc_url:
type: string
description: URL to relevant documentation
x-parser-schema-id:
request_id:
type: string
description: Unique identifier for this websocket connection
x-parser-schema-id:
x-parser-schema-id: STTErrorResponse
title: Error Response
description: Error information for STT WebSocket connections.
example: |-
{
"type": "error",
"title": "Invalid model",
"message": "The model is not valid, make sure it is a valid model ID.",
"error_code": "model_not_found",
"doc_url": "https://docs.cartesia.ai/build-with-cartesia/stt/latest",
"status_code": 400,
"request_id": "2ff8af53-4d38-479d-8287-58940f01c701"
}
bindings: []
extensions:
- id: x-parser-unique-object-id
value: sttErrorResponse
bindings: []
extensions: *ref_0
sendOperations:
- *ref_1
receiveOperations:
- *ref_2
sendMessages:
- *ref_3
- *ref_4
- *ref_5
receiveMessages:
- *ref_6
- *ref_7
- *ref_8
- *ref_9
- *ref_10
- *ref_11
- *ref_12
extensions:
- id: x-parser-unique-object-id
value: /stt/turns/websocket
securitySchemes:
- id: apiKey
name: X-API-Key
type: httpApiKey
description: API key passed in a header.
in: header
extensions: []
- id: accessTokenQuery
name: access_token
type: httpApiKey
description: >
A short-lived access token passed in a query param to make API requests
from a client.
This is particularly useful in the browser, where WebSockets do not
support headers.
See [Authenticate client
apps](/get-started/authenticate-your-client-applications) to generate an
access token.
in: query
extensions: []
````