---
openapi: 3.1.0
components:
  schemas:
    Language:
      type: string
      enum:
      - English
      - Hindi
      - Bengali
      - Marathi
      - Telugu
      - Tamil
      - Gujarati
      - Urdu
      - Bhojpuri
      - Kannada
      - Odia
      - Malayalam
      - Punjabi
      - Chhattisgarhi
      - Assamese
      - Maithili
      - Magahi
      - Santali
      - Kashmiri
      - Kashmiri_Arabic
      - Nepali
      - Sindhi
      - Sindhi_Arabic
      - Dogri
      - Konkani
      - Manipuri
      - Manipuri_Bengali
      - Bodo
      - Sanskrit
      - Auto
    NormalizeInput:
      type: object
      required:
      - text
      - language
      description: Input needed for Text Normalization.
      properties:
        text:
          type: string
          examples:
          - "While docking preparations are underway, SDX01 and SDX02 are 105 meters\
            \ apart in orbit as of 12 January IST 3:10 a.m."
          description: Unicode text which needs to be normalized.
        language:
          $ref: '#/components/schemas/Language'
          type: string
          examples:
          - English
          description: Language (Indic/English) of input-text
    Script:
      type: string
      enum:
      - Devanagari
      - Bengali
      - Gurmukhi
      - Gujarati
      - Odia
      - Tamil
      - Telugu
      - Kannada
      - Malayalam
      - English
      - Urdu
      - OlChiki
      - Meitei
    TextToSpeechInput:
      type: object
      required:
      - text
      - language
      - voiceName
      description: Input needed for synthesizing speech.
      properties:
        text:
          type: string
          examples:
          - "सौर मंडल में सूर्य और वह खगोलीय पिंड सम्मलित हैं, जो इस मंडल में एक द\
            ूसरे से गुरुत्वाकर्षण बल द्वारा बंधे हैं।"
          description: Unicode text for which speech needs to be synthesized.
        language:
          $ref: '#/components/schemas/Language'
          type: string
          examples:
          - Hindi
          description: Language (Indic/English) of input-text
        translateToLanguage:
          $ref: '#/components/schemas/Language'
          type: string
          examples:
          - Kannada
          description: "Optional. If provided, the input text is first translated\
            \ into this language and then synthesized into speech in the same language."
        voiceName:
          type: string
          examples:
          - Female1
          description: "Voice Name: Female1, Male1 etc."
        voiceStyle:
          $ref: '#/components/schemas/VoiceStyle'
          type: string
          examples:
          - Neutral
          description: "Voice Style: Tone or speaking manner of the generated voice."
          default: Neutral
        speechRate:
          type: number
          format: float
          examples:
          - 1.25
          description: "Control the speed of the generated speech. 1.0 is normal speed,\
            \ less than 1.0 is slower, greater than 1.0 is faster."
          default: 1.0
    TranslateInput:
      type: object
      required:
      - inputText
      - inputLanguage
      - outputLanguage
      description: Input needed for Transliteration.
      properties:
        inputText:
          type: string
          examples:
          - Chandrayaan-3 is the third and most recent lunar Indian Space Research
            exploration mission under the Chandrayaan programme of ISRO. It consists
            of a lander named Vikram and a rover named Pragyan similar to Chandrayaan-2.
            Its propulsion module behaves like an Orbiter. The propulsion module carries
            the lander and rover configuration until the spacecraft is in a 100-kilometre
            (62 mi) lunar orbit.
          description: Unicode text which needs to be transliterated.
        inputLanguage:
          $ref: '#/components/schemas/Language'
          type: string
          examples:
          - English
          description: Source language (Indic/English) of input-text
        outputLanguage:
          $ref: '#/components/schemas/Language'
          type: string
          examples:
          - Hindi
          description: Target language (Indic/English) in to which the inputText needs
            to be translated to
    TransliterateInput:
      type: object
      required:
      - inputText
      - outputScript
      description: Input needed for Transliteration.
      properties:
        inputText:
          type: string
          examples:
          - भारत का चंद्रयान-3 मिशन के तहत भेजा गया लैंडर चंद्रमा के दक्षिणी ध्रुव
            की सतह पर सफलतापूर्वक उतर गया।
          description: Unicode text which needs to be transliterated.
        inputScript:
          $ref: '#/components/schemas/Script'
          type: string
          examples:
          - Devanagari
          description: Script in which input text is given
        inputLanguage:
          $ref: '#/components/schemas/Language'
          type: string
          examples:
          - Hindi
          description: Language of input text
        outputScript:
          $ref: '#/components/schemas/Script'
          type: string
          examples:
          - Kannada
          description: Script in which the output text is expected
    VoiceStyle:
      type: string
      enum:
      - Neutral
      - News
      - Conversational
      - Book
      - Command
      - Happy
      - Sad
      - Fear
      - Disgust
      - Anger
      - Surprise
      - Custom
      description: "Voice Style: Tone or speaking manner of the generated voice. Possible\
        \ values: Neutral, Conversational, Happy, Sad, News, Command, etc."
  parameters:
    X-API-KEY:
      description: Specify your API key (consisting of 32 characters) for going beyond
        rate limits and/or getting access to premium voices
      in: header
      name: X-API-KEY
      schema:
        type: string
        maxLength: 32
    version:
      description: API version (v1 or v2)
      in: path
      name: version
      required: true
      schema:
        enum:
        - v1
        - v2
        default: v2
  securitySchemes:
    IdToken:
      type: http
      scheme: bearer
      bearerFormat: JWT
    JWT:
      type: http
      scheme: bearer
      bearerFormat: JWT
info:
  description: "These APIs provide access to Bhashini.ai Text-to-Speech (TTS), Speech-to-Text\
    \ (STT), Translation and OCR services"
  title: Bhashini.ai REST APIs
  version: 1.0.0
paths:
  /authorize:
    post:
      responses:
        "200":
          description: OK
      security:
      - IdToken: []
  /deepdub/detect-dialogs/{jobId}:
    get:
      parameters:
      - name: jobId
        in: path
        required: true
        schema:
          type: string
      - name: includeSession
        in: query
        schema:
          type: boolean
      - $ref: '#/components/parameters/X-API-KEY'
      responses:
        "200":
          description: OK
  /deepdub/detect-dialogs/{jobId}/events:
    get:
      parameters:
      - name: jobId
        in: path
        required: true
        schema:
          type: string
      - name: includeSession
        in: query
        schema:
          type: boolean
      - $ref: '#/components/parameters/X-API-KEY'
      responses:
        "200":
          description: OK
  /deepdub/exports/{jobId}:
    get:
      parameters:
      - name: jobId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      responses:
        "200":
          description: OK
  /deepdub/exports/{jobId}/artifact:
    get:
      parameters:
      - name: jobId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      responses:
        "200":
          description: OK
  /deepdub/sessions/get-or-create:
    post:
      summary: Get or create DeepDub session by user + source + video identity
      parameters:
      - $ref: '#/components/parameters/X-API-KEY'
      requestBody:
        content:
          application/json:
            schema:
              type: string
        required: true
      responses:
        "200":
          description: Resolved DeepDub session
          content:
            application/json: {}
  /deepdub/sessions/{sessionId}:
    get:
      parameters:
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      responses:
        "200":
          description: OK
    delete:
      parameters:
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      responses:
        "200":
          description: OK
  /deepdub/sessions/{sessionId}/artifacts/{artifactKey}:
    get:
      parameters:
      - name: artifactKey
        in: path
        required: true
        schema:
          type: string
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      responses:
        "200":
          description: OK
  /deepdub/sessions/{sessionId}/collaborators:
    delete:
      parameters:
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - name: emailId
        in: query
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      responses:
        "200":
          description: OK
    get:
      parameters:
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      responses:
        "200":
          description: OK
    post:
      parameters:
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      requestBody:
        content:
          application/json:
            schema:
              type: string
        required: true
      responses:
        "200":
          description: OK
  /deepdub/sessions/{sessionId}/detect-dialogs:
    post:
      parameters:
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      requestBody:
        content:
          multipart/form-data:
            schema:
              type: object
              properties:
                file:
                  type: object
                sourceLanguage:
                  type: string
                rangeStartSec:
                  type: string
                rangeEndSec:
                  type: string
                minDialogSec:
                  type: string
          application/json:
            schema:
              type: string
        required: true
      responses:
        "200":
          description: OK
  /deepdub/sessions/{sessionId}/dialogs:
    put:
      parameters:
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      requestBody:
        content:
          application/json:
            schema:
              type: string
        required: true
      responses:
        "200":
          description: OK
  /deepdub/sessions/{sessionId}/dialogs/{dialogId}:
    put:
      parameters:
      - name: dialogId
        in: path
        required: true
        schema:
          type: string
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      requestBody:
        content:
          application/json:
            schema:
              type: string
        required: true
      responses:
        "200":
          description: OK
    delete:
      parameters:
      - name: dialogId
        in: path
        required: true
        schema:
          type: string
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      responses:
        "200":
          description: OK
  /deepdub/sessions/{sessionId}/dialogs/{dialogId}/clip:
    get:
      parameters:
      - name: dialogId
        in: path
        required: true
        schema:
          type: string
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - name: targetLanguage
        in: query
        schema:
          $ref: '#/components/schemas/Language'
      - $ref: '#/components/parameters/X-API-KEY'
      responses:
        "200":
          description: OK
  /deepdub/sessions/{sessionId}/dialogs/{dialogId}/merge:
    post:
      parameters:
      - name: dialogId
        in: path
        required: true
        schema:
          type: string
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      requestBody:
        content:
          application/json:
            schema:
              type: string
        required: true
      responses:
        "200":
          description: OK
  /deepdub/sessions/{sessionId}/dialogs/{dialogId}/split:
    post:
      parameters:
      - name: dialogId
        in: path
        required: true
        schema:
          type: string
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      requestBody:
        content:
          application/json:
            schema:
              type: string
        required: true
      responses:
        "200":
          description: OK
  /deepdub/sessions/{sessionId}/export:
    post:
      parameters:
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      requestBody:
        content:
          application/json:
            schema:
              type: string
        required: true
      responses:
        "200":
          description: OK
  /deepdub/sessions/{sessionId}/extract-audio:
    post:
      parameters:
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      requestBody:
        content:
          multipart/form-data:
            schema:
              type: object
              properties:
                file:
                  type: object
                rangeStartSec:
                  type: string
                rangeEndSec:
                  type: string
          application/json:
            schema:
              type: string
        required: true
      responses:
        "200":
          description: OK
  /deepdub/sessions/{sessionId}/preferences:
    post:
      parameters:
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      requestBody:
        content:
          application/json:
            schema:
              type: string
        required: true
      responses:
        "200":
          description: OK
  /deepdub/sessions/{sessionId}/synthesize:
    post:
      parameters:
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      requestBody:
        content:
          application/json:
            schema:
              type: string
        required: true
      responses:
        "200":
          description: OK
  /deepdub/sessions/{sessionId}/synthesize/track:
    post:
      parameters:
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      requestBody:
        content:
          application/json:
            schema:
              type: string
        required: true
      responses:
        "200":
          description: OK
  /deepdub/sessions/{sessionId}/synthesize/{dialogId}:
    post:
      parameters:
      - name: dialogId
        in: path
        required: true
        schema:
          type: string
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      requestBody:
        content:
          application/json:
            schema:
              type: string
        required: true
      responses:
        "200":
          description: OK
  /deepdub/sessions/{sessionId}/transcribe:
    post:
      parameters:
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      requestBody:
        content:
          application/json:
            schema:
              type: string
        required: true
      responses:
        "200":
          description: OK
  /deepdub/sessions/{sessionId}/translate:
    post:
      parameters:
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      requestBody:
        content:
          application/json:
            schema:
              type: string
        required: true
      responses:
        "200":
          description: OK
  /deepdub/sessions/{sessionId}/transliterate:
    post:
      parameters:
      - name: sessionId
        in: path
        required: true
        schema:
          type: string
      - $ref: '#/components/parameters/X-API-KEY'
      requestBody:
        content:
          application/json:
            schema:
              type: string
        required: true
      responses:
        "200":
          description: OK
  /{version}/asr:
    post:
      summary: Run ASR
      description: Run ASR on the uploaded audio.
      parameters:
      - $ref: '#/components/parameters/X-API-KEY'
      - $ref: '#/components/parameters/version'
      requestBody:
        content:
          multipart/form-data:
            schema:
              type: object
              properties:
                file:
                  type:
                  - object
                  - string
                  format: binary
                language:
                  $ref: '#/components/schemas/Language'
                  description: Language of input-audio
        required: true
      responses:
        default:
          description: Recognized Text
          content:
            application/json: {}
        "200":
          description: Recognized text
        "400":
          description: Invalid audio file or unsupported language
        "401":
          description: Missing/Invalid X-API-KEY or JWT
  /{version}/normalize:
    post:
      summary: Indic Text Normalizer
      description: Expand numbers and abbreviations in English or Indian language
        text
      parameters:
      - $ref: '#/components/parameters/X-API-KEY'
      - $ref: '#/components/parameters/version'
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/NormalizeInput'
        required: true
      responses:
        default:
          description: Normalized text
          content:
            text/plain: {}
        "400":
          description: Invalid text/language in the request body
        "401":
          description: Missing/Invalid X-API-KEY or JWT
  /{version}/ocr:
    post:
      summary: Run Cloud OCR
      description: Run Cloud OCR on the uploaded document/image.
      parameters:
      - $ref: '#/components/parameters/X-API-KEY'
      - $ref: '#/components/parameters/version'
      requestBody:
        content:
          multipart/form-data:
            schema:
              type: object
              properties:
                file:
                  type:
                  - object
                  - string
                  format: binary
        required: true
      responses:
        default:
          description: Recognized Text
          content:
            application/xml: {}
        "400":
          description: Invalid file in the request body
        "401":
          description: Missing/Invalid X-API-KEY or JWT
  /{version}/synthesize:
    post:
      summary: Synthesize speech from text
      description: Use Bhashini.ai TTS to synthesize speech from Unicode text
      parameters:
      - $ref: '#/components/parameters/X-API-KEY'
      - $ref: '#/components/parameters/version'
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/TextToSpeechInput'
        required: true
      responses:
        default:
          description: Synthesized speech
          content:
            audio/mpeg: {}
        "400":
          description: Invalid language/voiceName/text in the request body
        "401":
          description: Missing/Invalid X-API-KEY or JWT
  /{version}/translate:
    post:
      summary: "Translate text: English->Indic or Indic->English or Indic->Indic"
      description: Translate text from English to Indian language or Indian language
        to English or from one Indian language to another.
      parameters:
      - $ref: '#/components/parameters/X-API-KEY'
      - $ref: '#/components/parameters/version'
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/TranslateInput'
        required: true
      responses:
        default:
          description: Text translated into target language
          content:
            text/plain: {}
        "400":
          description: Invalid inputText/inputLanguage/outputLanguage in the request
            body
        "401":
          description: Missing/Invalid X-API-KEY or JWT
  /{version}/transliterate:
    post:
      summary: Transliterate Indic text
      description: See Indic text in desired script
      parameters:
      - $ref: '#/components/parameters/X-API-KEY'
      - $ref: '#/components/parameters/version'
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/TransliterateInput'
        required: true
      responses:
        default:
          description: Text transliterated into target script
          content:
            text/plain: {}
        "400":
          description: Invalid inputText/inputScript/outputScript in the request body
servers:
- url: https://tts.bhashini.ai