How to set up a custom vocabulary for better transcription accuracy?

Custom Vocabulary for Transcription

To achieve higher transcription accuracy—especially when working with specialized terms, brand names, or recurring phrases—you can enable the custom_vocabulary feature. This lets you guide the transcription engine by telling it which words matter most, how they should be pronounced, and in which language.


{
  "audio_url": "YOUR_AUDIO_URL",
  "custom_vocabulary": true,
  "custom_vocabulary_config": {
    "vocabulary": [
      "Westeros",
      {
        "value": "Stark"
      },
      {
        "value": "Night's Watch",
        "pronunciations": ["Nightz Vatch"],
        "intensity": 0.4,
        "language": "de"
      }
    ],
    "default_intensity": 0.6
  }
}

💡 Note: Using a string in place of an object is equivalent to:


{
	"value": "string"
}

⚙️ Configuration Options

Field	Required	Description	Range / Default
`default_intensity`	❌	Global intensity of the feature.	`0.0–1.0`, default `0.5`
`vocabulary`	✅	Array of strings and/or objects
`vocabulary.value`	✅	The text to match and prioritize in transcription.	—
`vocabulary.pronunciations`	❌	Pronunciations of the word, either in transcription language or `vocabulary.language`.	—
`vocabulary.intensity`	❌	Intensity for this specific word.	`0.0–1.0`, default `0.5`
`vocabulary.language`	❌	Language in which the word is pronounced for sound comparison.	Default: transcription language

To learn more about the custom vocabulary feature, you can check our doc : https://docs.gladia.io/chapters/live-stt/features#custom-vocabulary