Kensho Scribe Batch API v2

Asynchronous audio and video transcription REST API. POST a media file to start a transcription job, then poll the transcription ID for completion. Optimised for finance and business audio (earnings calls, conferences). Companion real-time WebSocket API and Human-in-the-Loop batch option are documented alongside.

Kensho Scribe Batch API v2 is one of 8 APIs that S&P Global publishes on the APIs.io network, described by a machine-readable OpenAPI specification.

This API exposes 1 machine-runnable capability that can be deployed as REST, MCP, or Agent Skill surfaces via Naftiko.

Tagged areas include Speech to Text, Transcription, Audio, Video, and Batch. The published artifact set on APIs.io includes an OpenAPI specification, API documentation, an API reference, authentication docs, a JSON-LD context, and 1 Naftiko capability spec.

OpenAPI Specification

kensho-scribe-batch-v2-openapi.yml Raw ↑
openapi: 3.0.0
servers:
- url: https://scribe.kensho.com
info:
  version: 2.0.0
  title: Scribe Batch API
  description: 'Scribe''s Batch API is a RESTful API allowing users to asynchronously transcribe audio or video files.

    '
security:
- BearerAuth: []
paths:
  /api/v2/transcription:
    post:
      summary: Start A New Transcription
      operationId: startTranscription
      description: "Starts a transcription request for a single audio or video file.\n\nWe support a wide array of audio formats\
        \ along with their respective compression / codec\noptions including:\n - AAC\n - FLAC\n - MP1\n - MP3\n - WAV\n\n\
        Currently we only support MP4 video files where the audio streams _must_ be one\nof the supported audio formats.\n\
        \nFiles must be less than or equal to 1 gigabyte in size, and the upload must\ntake less than 30 minutes.\n\nMultichannel\
        \ audio will be merged into a single stream prior to transcription.\nIn very rare cases this can cause transcription\
        \ issues with stereo audio when\nthe two channels have near identical content that is slightly out of phase.\n"
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/jsonOptions'
          multipart/form-data:
            schema:
              $ref: '#/components/schemas/mpfdOptions'
            encoding:
              media:
                contentType: application/mp3, audio/aac, audio/flac, audio/m4a, audio/mp3, audio/mpeg, audio/mpeg3, audio/wav,
                  audio/wave, audio/x-flac, audio/x-m4a, audio/x-mpeg-3, audio/x-wav, video/mp4
      callbacks:
        transcription_complete:
          callback_uri:
            put:
              summary: An optional callback invoked upon transcription completion.
              description: 'If the request includes the `callback_uri` option then this endpoint will be called when the transcription
                is complete.

                The endpoint must be externally accessible and any response other than 200 will be seen as failure.  A failure
                to call

                the callback will not result in a failed transcription.

                '
              requestBody:
                required: true
                content:
                  application/json:
                    schema:
                      $ref: '#/components/schemas/callback'
              responses:
                '200':
                  description: The callback was successfully received.
      responses:
        '201':
          description: Transcription was successfully started.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/transcriptionId'
        '400':
          description: An invalid request was made to the server.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/error'
        '401':
          description: 'Authentication information was not provided - check to ensure that `Authorization`

            header is present.

            '
        '404':
          description: 'The transcript, for the `transcription_id`, cannot be found.

            '
        '429':
          description: 'The request was rate limited.  Limit the requests to a maximum of 5 per second to avoid

            getting this.

            '
  /api/v2/transcription/{transcription_id}:
    head:
      summary: Query a transcription by ID
      operationId: queryTranscription
      description: Determines if a transcript is available for download or not.
      parameters:
      - in: path
        name: transcription_id
        schema:
          type: string
        required: true
        description: ID of a transcription to query for
      responses:
        '200':
          description: 'The transcription, for the `transcription_id` is complete and can be downloaded

            for the returned `Content-Type`.

            '
          headers:
            Expires:
              schema:
                type: string
                format: date-time
              description: The date and time, after which, the transcription will no longer be retrievable
        '202':
          description: 'The transcription, for the `transcription_id` is still in progress.  Retrying

            later will yield a successful result.

            '
        '400':
          description: An invalid request was made to the server.
        '401':
          description: 'Authentication information was not provided - check to ensure that `Authorization`

            header is present.

            '
        '404':
          description: 'The transcript, for the `transcription_id`, cannot be found.

            '
        '406':
          description: 'A transcription cannot be queried for the given ''Accept'' mime type.

            '
        '429':
          description: 'The request was rate limited.  Limit the requests to a maximum of 5 per second to avoid

            getting this.

            '
    get:
      summary: Download A Transcription By ID
      operationId: downloadTranscription
      description: Get the resulting transcript from the transcription request.
      parameters:
      - in: path
        name: transcription_id
        schema:
          type: string
        required: true
        description: ID of a transcription to retrieve
      responses:
        '200':
          description: Results of a transcription request
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/jsonTranscript'
              example:
                transcript: Hello world
                accuracy: 0.99
                slice_meta:
                - transcript: Hello world
                  accuracy: 0.99
                  start_ms: 0
                  duration_ms: 2000
                  speaker_id: 0
                  speaker_accuracy: 1
                  token_meta:
                  - transcript: Hello
                    accuracy: 0.99
                    start_ms: 0
                    duration_ms: 1000
                    align_success: true
                  - transcript: world
                    accuracy: 0.99
                    start_ms: 1000
                    duration_ms: 1000
                    align_success: true
            text/vtt:
              schema:
                $ref: '#/components/schemas/vttTranscript'
              example: 'WEBVTT


                NOTE

                Confidence: 99%


                00:00:00.000 --> 00:00:2.000

                <v speaker_0>Hello world</v>

                '
            application/vnd.openxmlformats-officedocument.wordprocessingml.document:
              schema:
                description: A Microsoft Word document
                type: string
                format: binary
          headers:
            Expires:
              schema:
                type: string
                format: date-time
              description: The date and time, after which, the transcription will no longer be retrievable
        '202':
          description: 'The transcription, with the `transcription_id` is still in progress.  Retrying

            later will yield a successful result.

            '
        '400':
          description: An invalid request was made to the server.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/error'
        '401':
          description: 'Authentication information was not provided - check to ensure that `Authorization`

            header is present.

            '
        '404':
          description: 'The transcript, for the `transcription_id`, cannot be found.

            '
        '406':
          description: 'A transcription cannot be returned for the given ''Accept'' mime type.

            '
        '429':
          description: 'The request was rate limited.  Limit the requests to a maximum of 5 per second to avoid

            getting this.

            '
    delete:
      summary: Delete All Transcription Data By ID
      operationId: deleteTranscription
      description: Attempts to ensure that all stored data for a transcription is no longer available on the system.
      parameters:
      - in: path
        name: transcription_id
        schema:
          type: string
        required: true
        description: ID of a transcription to delete
      responses:
        '200':
          description: "All stored data associated with the transcription is no longer available.\nThis could mean that:\n\
            \ * The transcription data existed and is no longer available.\n * The transcription data did not exist and is\
            \ still no longer available.\n * The transcription data was not available to you and so, as far as you are concerned,\
            \ the data is no longer available.\n"
        '400':
          description: An invalid request was made to the server.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/error'
        '401':
          description: 'Authentication information was not provided - check to ensure that `Authorization`

            header is present.

            '
        '429':
          description: 'The request was rate limited.  Limit the requests to a maximum of 5 per second to avoid

            getting this.

            '
components:
  securitySchemes:
    BearerAuth:
      type: http
      scheme: bearer
      bearerFormat: JWT
      description: 'A Kensho access token is required for all transcription requests.  For more information on access tokens,

        and how to obtain them, view the [authentication guide](../../authentication).

        '
  schemas:
    commonOptions:
      type: object
      properties:
        callback_uri:
          type: string
          format: uri
          description: 'A URI (currently restricted to http/https) which Scribe can send information to when the

            transcription is complete.


            When the URI protocol is http or https we will make a PUT request to the callback_uri to

            indicate a change in the status of the transcription job.  The PUT request body will be

            JSON and contain the transcription id and if the transcription completed successfully or not:

            ```json

            { ''transcription_id'': ''<ID>'', ''result'': ''success'' }

            ```


            A separate call can then be made to retrieve the transcript (or any failures while generating it).

            '
        transcriber:
          type: string
          enum:
          - human
          - machine
          default: machine
          description: 'The type of transcription to perform - either machine only (AI) or human assisted (human in the loop).

            '
        media_language:
          type: string
          enum:
          - cmn
          - da
          - dan
          - de
          - deu
          - en
          - eng
          - es
          - fr
          - fra
          - it
          - ita
          - ja
          - jpn
          - ko
          - kor
          - nl
          - nld
          - 'no'
          - nor
          - por
          - pt
          - ru
          - rus
          - spa
          - yue
          - zh
          - zho
          default: en
          description: 'The source language of the media.  If the media is not in english then the contents will have to be
            translated

            prior to transcription.  This is a single, ISO-639, language code which can be two characters for ISO-639-1 (ie
            en)

            or three characters for ISO-639-3 (ie zho).


            **NOTE - Only applicable when `human` is specified as the transcriber.**

            '
        priority:
          type: string
          enum:
          - low
          - medium
          - high
          default: medium
          description: 'This is the priority of the transcription and is dictated by contractual turn around times.


            **NOTE - Only applicable when `human` is specified as the transcriber.**

            '
        context:
          oneOf:
          - type: string
            description: 'a single or multiline block of text with ''context'' on this transcription.

              '
            example: Q1 Earnings Call
          - type: object
            additionalProperties: true
            properties:
              id:
                type: string
                description: 'A short ID to pair with the transcription - typically something to cross reference the

                  transcription with.

                  '
                example: abc_internal_id
              title:
                type: string
                description: 'A (typically) one line title / name to associate with the transcription.

                  '
                example: Q1 earnings
              description:
                type: string
                description: 'A single or multiline block of text with more descriptive information about the transcription

                  '
                example: 'Q1 earnings calls featuring Jane the CEO, with guest speaker John.

                  '
            description: TODO
          description: 'Additional ''context'' information that can be attached to the transcription.  This is typically never
            read

            on the Kensho side, but it can be used to recall or cross-reference internal information about the

            transcript.  A typical scenario could be one where transcriptions are submitted with titles, names and

            tracked with internal IDs which is needed when retrieving the transcript.

            '
    jsonOptions:
      allOf:
      - type: object
        required:
        - media_url
        properties:
          media_url:
            type: string
            format: url
            example: https://example.com/earnings_report.mp3
            description: 'The URL pointing to the media to transcribe.

              This needs to be a full URL, less than 2048 characters, where Scribe can GET the data from without any other

              intervention (credentials, tokens, etc.).

              If Scribe cannot GET from that URL then the job will fail.

              '
      - $ref: '#/components/schemas/commonOptions'
    mpfdOptions:
      type: object
      required:
      - media
      properties:
        media:
          type: string
          format: binary
          description: 'The binary data for the media to transcribe.

            Along with the binary data you will want to provide the `filename` in the value for `Content-Disposition`,

            and the mime type in the value for `Content-Type`.  The `Content-Type` currently can be one of:

            - application/mp3

            - audio/aac

            - audio/flac

            - audio/m4a

            - audio/mp3

            - audio/mpeg

            - audio/mpeg3

            - audio/wav

            - audio/wave

            - audio/x-flac

            - audio/x-m4a

            - audio/x-mpeg-3

            - audio/x-wav

            - video/mp4

            '
        options:
          $ref: '#/components/schemas/commonOptions'
    error:
      type: object
      properties:
        detail:
          type: string
          description: A detailed statement about why the error occurred.
          example: Specified content type cannot be returned
        msg_code:
          type: string
          description: An error code which can be used to get more information about the problem.
          example: S2T-BA-0010
    callback:
      type: object
      properties:
        transcription_id:
          type: string
          description: The ID of the transcription
          example: 76fc238be38d49cda51dd396ca69215d
        result:
          type: string
          description: The final result of the transcription - one of 'success' or 'failure'
          example: success
      required:
      - transcription_id
      - result
    transcriptionId:
      type: object
      properties:
        transcription_id:
          type: string
          description: The ID of the transcription
          example: 76fc238be38d49cda51dd396ca69215d
      required:
      - transcription_id
    vttTranscript:
      description: Transcribed video text tracks
      type: string
    jsonTranscript:
      description: A structured output of the transcript - ideal for programmatic or custom uses of transcriptions
      type: object
      properties:
        transcript:
          description: The full text of the transcript
          type: string
        accuracy:
          type: number
          minimum: 0
          maximum: 1
          description: 'An indicator of accuracy - a higher number gives a higher confidence towards the transcription.

            '
        slice_meta:
          type: array
          description: Array of slices of the transcribed audio
          items:
            $ref: '#/components/schemas/sliceMeta'
    sliceMeta:
      description: A single slice of transcribed audio
      type: object
      properties:
        transcript:
          type: string
          description: The transcribed text of the slice
        accuracy:
          type: number
          minimum: 0
          maximum: 1
          description: 'An indicator of the accuracy of the transcribed slice - a higher number gives a higher confidence.

            '
        start_ms:
          type: integer
          description: The start time, from the beginning of the file, of the slice in milliseconds
        duration_ms:
          type: integer
          description: The duration of the slice in milliseconds
        speaker_id:
          type: integer
          description: 'The ID of the speaker in the stream of audio.  Typically this increments from 0 (first detected speaker)

            and, when negative (unknown speaker), indicates that we are not confident in saying we''ve heard this

            speaker in the audio before.

            '
        speaker_accuracy:
          minimum: 0
          maximum: 1
          description: 'An indicator of the accuracy of the speaker identification - a higher number gives a higher confidence.

            '
        token_meta:
          type: array
          description: Array of tokens of the transcribed slice
          items:
            $ref: '#/components/schemas/tokenMeta'
    tokenMeta:
      description: A single transcribed token
      type: object
      properties:
        transcript:
          type: string
          description: The transcribed text of the token
        accuracy:
          type: number
          minimum: 0
          maximum: 1
          description: 'An indicator of the accuracy of the transcribed token - a higher number gives a higher confidence.

            '
        start_ms:
          type: integer
          description: The token start time, in milliseconds, from the beginning of the file.
        duration_ms:
          type: integer
          description: The token duration, in milliseconds.
        align_success:
          type: boolean
          description: Indicates whether the token has a definitive timestamp