Skip to main content
POST
Speech transcription

Authorizations

Authorization
string
header
required

Bearer authentication header of the form Bearer <token>, where <token> is your auth token.

Body

multipart/form-data
file
file
required

Audio file to be transcribed.

model
file
required

Speech transcription model ID.

language
string

Enter the audio language, ISO 639-1 language codes are recommended.

Example:

"zh"

languages
string[]

Possible language codes.

Example:

"zh,en"

prompt
string

Prompt words to guide transcription style, terminology, and context.

Example:

"Meeting Minutes Terminology: Tikway"

keywords
string[]

A word or phrase that needs to be identified.

Example:

"Tikway,API"

response_format
string

Response format.

Example:

"json"

temperature
string

Sampling temperature.

Example:

"0"

stream
number

Whether to use SSE streaming returns.

Example:

"false"

chunking_strategy
boolean

Chunking strategy.

Example:

"auto"

known_speaker_names
string[]

Known speaker name.

Example:

"Alice,Bob"

known_speaker_references
string[]

Audio Base64 data URL of known speaker.

Example:

"data:audio/wav;base64,UklGRg=="

Response

200 - application/json

Non-streaming speech transcription JSON response. Compatible with ordinary transcription, detailed timestamp transcription and speaker separation results.

text
string
required

Transcribe the text completely.

task
string

Task type. Verbose JSON and speaker-detached responses typically return this field.

Allowed value: "transcribe"
language
string

Audio language detected or used.

duration
number

Audio duration, in seconds.

Required range: x >= 0
words
object[]

Word-level timestamp, returned only when word granularity is requested.

segments
object[]

Transcribe the results in segments. verbose_json returns timestamp segments; diarized_json also returns speaker information.

logprobs
object[]

May be returned when requesting include[]=logprobs.

usage
TokenUsage · object

The token or audio duration usage returned by the upstream.