vocabulary:
  name: Salad Transcription API Vocabulary
  description: >-
    Operational vocabulary for the Salad Transcription API covering speech-to-text
    concepts, transcription job lifecycle, audio/video formats, diarization,
    and output formats.
  version: "1.0"
  created: "2026-05-02"
  modified: "2026-05-02"
  tags:
    - Audio Transcription
    - Media Processing
    - Salad
    - Speech Recognition

  domains:
    - name: Transcription Concepts
      description: "Core terms for speech-to-text transcription."
      terms:
        - term: Transcription
          definition: >-
            The process of converting spoken audio or video content into written text.
          aliases: [Speech-to-Text, STT, ASR]
          related: [Segment, Diarization, Language Code]

        - term: Transcription Job
          definition: >-
            A submitted request to transcribe a media file. Has a lifecycle from
            pending through created, running, succeeded, or failed.
          aliases: [Job]
          related: [Job ID, Status, Output]

        - term: Job ID
          definition: >-
            A unique identifier returned when a transcription job is submitted.
            Used to retrieve results via the GET endpoint.
          aliases: [ID, Trace ID]
          related: [Transcription Job, Status]

        - term: Segment
          definition: >-
            A chunk of transcribed speech with start time, end time, text content,
            and optional speaker label.
          aliases: [Utterance, Chunk]
          related: [Word Segment, Speaker, Timestamp]

        - term: Word Segment
          definition: >-
            The finest unit of transcription output, providing start/end times
            and confidence score for each individual word.
          aliases: [Word-Level Timestamp, Word]
          related: [Segment, Score]

    - name: Configuration Options
      description: "Parameters controlling transcription behavior."
      terms:
        - term: Language Code
          definition: >-
            A BCP-47 language code specifying the spoken language to transcribe.
            97 languages supported. Defaults to 'en' (English) if not specified.
          aliases: [Language, Locale]
          related: [Transcription, Diarization]

        - term: Diarization
          definition: >-
            The process of separating audio into distinct speaker turns, labeling
            each segment with a speaker identifier (e.g., SPEAKER_00, SPEAKER_01).
            Requires specifying a supported language_code.
          aliases: [Speaker Separation, Speaker Identification]
          related: [Segment, Language Code]

        - term: Word-Level Timestamps
          definition: >-
            A transcription option that includes start and end time markers for
            every individual word in the output, in addition to segment-level timing.
          aliases: [Word Timestamps]
          related: [Segment, Word Segment]

        - term: SRT Output
          definition: >-
            SubRip Subtitle format. When enabled, the transcription output includes
            an srt_content field with properly formatted caption text for use in
            video players or subtitle tracks.
          aliases: [SRT, Subtitles, Captions]
          related: [Segment, Output]

    - name: Job Lifecycle
      description: "States in the transcription job workflow."
      terms:
        - term: Pending
          definition: >-
            Initial state before the job is picked up for processing.
          related: [Created, Running, Job]

        - term: Created
          definition: >-
            The job has been queued for processing.
          related: [Pending, Running, Job]

        - term: Running
          definition: >-
            The transcription is actively being processed on Salad's GPU network.
          related: [Created, Succeeded, Failed, Job]

        - term: Succeeded
          definition: >-
            The transcription completed successfully. The output field contains results.
          related: [Running, Output, Segment]

        - term: Failed
          definition: >-
            The transcription could not be completed after three automatic retries.
            Common causes: file over 3GB, file unavailable, or duration over 2.5 hours.
          related: [Running, Job]

    - name: Media Formats
      description: "Supported input media formats."
      terms:
        - term: Audio Formats
          definition: >-
            Supported audio file formats for transcription: AIFF, FLAC, M4A, MP3, WAV.
          related: [Transcription, URL]

        - term: Video Formats
          definition: >-
            Supported video file formats for transcription: MP4, MKV, MOV, WEBM, WMA.
          related: [Transcription, URL]

        - term: Media URL
          definition: >-
            The publicly accessible URL pointing to the audio or video file to
            be transcribed. Must be on a public server or storage bucket (e.g., S3).
            YouTube and Google Drive links are not supported.
          related: [Transcription Job, Input]