Speech-to-Text Conversion

Core engine for high-fidelity audio-to-text transcription.

<?xml version="1.0"?>
<query>
    <userid>[UserID]</userid> <!-- required -->
    <userkey>[UserKey]</userkey> <!-- required -->
    <action>[Action]</action> <!-- required -->
    <source>[SourceFile]</source> <!-- required -->
    <format>
        <output>closed_captions</output> <!-- required-->
        <stt> 
            <model>[base|tiny|small|medium|largev2|largev3]</model> 
            <language>[LanguageCode]</language> 
            <prompt></prompt> 
            <refine_timing>[yes|no]</refine_timing>
            <normalize_non_verbal>[Default|Remove|Parentheses|SquareBrackets]</normalize_non_verbal>
            <normalize_music>[MusicCueText]</normalize_music>
            <audio_stream>
                [AudioStreamDefinition]
            </audio_stream>
        </stt>
    </format>
</query>

{
  "query": {
    "userid": "[UserID]", // required
    "userkey": "[UserKey]", // required
    "action": "[Action]", // required
    "source": "[SourceFile]", // required
    "format": {
      "output": "closed_captions",
      "stt": {
        "model": "[base|tiny|small|medium|largev2|largev3]",
        "language": "[LanguageCode]",
        "prompt": "",
        "refine_timing": "[yes|no]",
        "normalize_non_verbal": "[Default|Remove|Parentheses|SquareBrackets]",
        "normalize_music": "[MusicCueText]",
        "audio_stream": {
          [AudioStreamDefinition]
        }
      }
    }
  }
}

Parameter	Description	Allowed Values	Default Value
model	Language model to use	base, tiny, small, medium, largev2, largev3 (experimental)	largev2
language	Language of speech	2 or 3 characters language code	none
prompt	Prompt for speech to text conversion	Text	none
refine_timing	Refine timing	yes, no	no
normalize_non_verbal	How non-verbal cues (e.g. sound effects) are formatted. `Default`: keep as produced by the model; `Remove`: strip them; `Parentheses`: wrap in `( )`; `SquareBrackets`: wrap in `[ ]`	Default, Remove, Parentheses, SquareBrackets	Default
normalize_music	Text/symbol used to denote music cues (case-preserved)	Text	none
audio_stream	Define audio stream to handle	see Audio Stream definition