swagger.yaml

basePath: /
definitions:
  gateway.VoiceLabel:
    properties:
      category:
        type: string
      label:
        type: string
    type: object
  models.VoiceType:
    enum:
    - system
    - cloned
    type: string
    x-enum-varnames:
    - VoiceTypeSystem
    - VoiceTypeCloned
  null.String:
    properties:
      string:
        type: string
      valid:
        type: boolean
    type: object
  tts.convertTextDto:
    properties:
      save:
        type: boolean
      text:
        type: string
      voiceId:
        type: integer
      voiceType:
        $ref: '#/definitions/models.VoiceType'
    type: object
  tts.voiceDto:
    properties:
      description:
        type: string
      disabled:
        type: boolean
      id:
        type: integer
      labels:
        items:
          $ref: '#/definitions/gateway.VoiceLabel'
        type: array
      name:
        type: string
      new:
        type: boolean
      previewUrl:
        allOf:
        - $ref: '#/definitions/null.String'
        description: '@deprecated'
      previewUrls:
        items:
          type: string
        type: array
      rating:
        type: number
      voiceTier:
        type: string
    type: object
  vc.voiceDto:
    properties:
      description:
        type: string
      disabled:
        type: boolean
      id:
        type: integer
      labels:
        items:
          $ref: '#/definitions/gateway.VoiceLabel'
        type: array
      name:
        type: string
      new:
        type: boolean
      previewUrl:
        allOf:
        - $ref: '#/definitions/null.String'
        description: '@deprecated'
      previewUrls:
        items:
          type: string
        type: array
      rating:
        type: number
      voiceTier:
        type: string
    type: object
  voice.publicVoiceDto:
    properties:
      description:
        type: string
      disabled:
        type: boolean
      id:
        type: integer
      imageKey:
        type: string
      labels:
        items:
          $ref: '#/definitions/gateway.VoiceLabel'
        type: array
      name:
        type: string
      new:
        type: boolean
      previewUrl:
        allOf:
        - $ref: '#/definitions/null.String'
        description: '@deprecated'
      previewUrls:
        items:
          type: string
        type: array
      rating:
        type: number
      voiceTier:
        type: string
    type: object
  voice_cloned.UpdateVoiceCloneDto:
    properties:
      name:
        type: string
    type: object
  voice_cloned.VoiceClonedDto:
    properties:
      createdAt:
        type: string
      id:
        type: integer
      imageUrl:
        type: string
      name:
        type: string
      previewUrls:
        items:
          type: string
        type: array
    type: object
host: api.gemelo.ai
info:
  contact: {}
  description: gemelo.ai API documentation
  title: gemelo.ai API
  version: 1.0.0
paths:
  /v1/cloned-voices:
    get:
      description: Returns all cloned voices
      parameters:
      - description: Max records per request
        in: query
        name: limit
        type: integer
      - description: Page indicator ('next' and 'prev' fields in the response)
        in: query
        name: cursor
        type: string
      responses:
        "200":
          description: OK
          schema:
            items:
              $ref: '#/definitions/voice_cloned.VoiceClonedDto'
            type: array
      security:
      - ClientKey: []
      - ApiKey: []
      tags:
      - Voice cloning
    post:
      consumes:
      - multipart/form-data
      description: Creates new cloned voice
      parameters:
      - description: .wav file (up to 60s / 32MB)
        in: formData
        name: audio
        required: true
        type: file
      - description: Cloned voice name
        in: formData
        name: name
        required: true
        type: string
      responses:
        "200":
          description: OK
          schema:
            $ref: '#/definitions/voice_cloned.VoiceClonedDto'
      security:
      - ClientKey: []
      - ApiKey: []
      tags:
      - Voice cloning
  /v1/cloned-voices/{clonedVoiceID}:
    delete:
      description: Deletes the cloned voice
      parameters:
      - description: Cloned Voice ID
        in: path
        name: clonedVoiceID
        required: true
        type: integer
      responses:
        "204":
          description: No Content
      security:
      - ClientKey: []
      - ApiKey: []
      tags:
      - Voice cloning
    patch:
      description: Updates the cloned voice
      parameters:
      - description: Cloned Voice ID
        in: path
        name: clonedVoiceID
        required: true
        type: integer
      - description: Voice cloning update input
        in: body
        name: body
        required: true
        schema:
          $ref: '#/definitions/voice_cloned.UpdateVoiceCloneDto'
      responses:
        "200":
          description: OK
          schema:
            $ref: '#/definitions/voice_cloned.VoiceClonedDto'
      security:
      - ClientKey: []
      - ApiKey: []
      tags:
      - Voice cloning
  /v1/tts/convert:
    post:
      description: Converts text to speech audio file
      parameters:
      - description: TTS input
        in: body
        name: body
        required: true
        schema:
          $ref: '#/definitions/tts.convertTextDto'
      responses:
        "200":
          description: Binary audio file
          schema:
            type: file
      security:
      - ClientKey: []
      - ApiKey: []
      tags:
      - Text to speech
  /v1/tts/voices:
    get:
      description: Returns all voices available for the Text to speech
      responses:
        "200":
          description: OK
          schema:
            items:
              $ref: '#/definitions/tts.voiceDto'
            type: array
      security:
      - ClientKey: []
      - ApiKey: []
      tags:
      - Text to speech
  /v1/vc/convert:
    post:
      consumes:
      - multipart/form-data
      description: Converts voice in the uploaded audio file
      parameters:
      - description: .wav file to convert
        in: formData
        name: file
        required: true
        type: file
      - description: VoiceID
        in: query
        name: voiceId
        required: true
        type: integer
      - description: Voice type (system/cloned)
        in: query
        name: voiceType
        type: string
      responses:
        "200":
          description: Binary audio file
          schema:
            type: file
      security:
      - ClientKey: []
      - ApiKey: []
      tags:
      - Voice conversion
  /v1/vc/voices:
    get:
      description: Returns all voices available for the Voice Conversion
      responses:
        "200":
          description: OK
          schema:
            items:
              $ref: '#/definitions/vc.voiceDto'
            type: array
      security:
      - ClientKey: []
      - ApiKey: []
      tags:
      - Voice conversion
  /v1/voices:
    get:
      description: Returns all system voices available for TTS and VC
      responses:
        "200":
          description: OK
          schema:
            items:
              $ref: '#/definitions/voice.publicVoiceDto'
            type: array
      security:
      - ClientKey: []
      - ApiKey: []
      tags:
      - Voice conversion
schemes:
- https
securityDefinitions:
  ApiKey:
    in: header
    name: X-API-Key
    type: apiKey
  ClientKey:
    in: header
    name: X-Client-Key
    type: apiKey
swagger: "2.0"