Skip to main content

Documentation Index

Fetch the complete documentation index at: https://vowen.mintlify.app/llms.txt

Use this file to discover all available pages before exploring further.

Custom Vocabulary

Add words and phrases that the transcription model should recognize. This is especially useful for:
  • Product names (Kubernetes, PostgreSQL, Figma)
  • People’s names (unusual spellings)
  • Industry jargon and acronyms
  • Company-specific terminology

Adding Words

  1. Open the Vowen app and click Dictionary in the left sidebar
  2. The Dictionary tab is selected by default (you’ll also see Threads and Expansions tabs)
  3. Click + Add Word in the top right
  4. Type your word or phrase and save

How It Works

Custom vocabulary words are provided as “hints” to the transcription model. This helps the model recognize these terms during speech-to-text processing, improving accuracy for domain-specific language.
Custom vocabulary works with most local models that support prompting (Small and larger). It may not have an effect on Tiny or cloud models that don’t support vocabulary hints.

Dictionary vs Threads

The same Dictionary page in the sidebar also hosts a Threads tab. Threads are a separate feature: text replacement rules that swap one phrase for another after transcription completes. Use Dictionary when the model can’t recognize a word at all; use Threads when it recognizes the word but spells it wrong.
FeatureDictionary (Vocabulary)Threads (Snippets)
PurposeHelp the model recognize a wordReplace one phrase with another
When appliedDuring transcriptionAfter transcription
ExampleAdd “Kubernetes” so it’s recognizedReplace “cube nets” with “Kubernetes”
Use whenModel doesn’t recognize a term at allModel recognizes the term but spells it wrong

Tips

  • Add the exact spelling you want (the model will try to match it)
  • Include both the full form and abbreviation if relevant
  • You don’t need to add common English words, only specialized terms
Pro tip: If a word is consistently misspelled even after adding it to the Dictionary, create a Thread replacement as a backup. The Dictionary nudges the model in the right direction during transcription; the Thread guarantees the final output is correct.