Skip to main content

Custom Vocabulary

Account custom vocabulary is a list of domain-specific terms the transcriber should recognize. Add brand names, product names, jargon, or unusual proper nouns; future transcriptions pick them up. It lives under Settings → Other → Account custom vocabulary.

Adding terms​

  1. Open Settings → Other → Account custom vocabulary.
  2. Type a term in the input box.
  3. Press Enter to add it. Repeat for each term.

You can add up to 100 terms. The list applies workspace-wide.

What to add​

Good candidates:

  • Brand names. Your company, your competitors, partners.
  • Product names. Internal SKUs, codenames, abbreviations.
  • Industry jargon. Acronyms and terms specific to your field.
  • Unusual person names. Board members, executives, frequent interviewees.
  • Place names. Cities or regions the model might mishear.

Skip common English words; the underlying model handles those.

Tips​

  • One term per entry. "User Evaluation" is one term; don't split it across two entries.
  • Spell exactly as you want it transcribed. "GPT-4" vs "GPT 4" gives different output.
  • Casing matters for proper nouns. "Kafka" (the queue) and "kafka" (lowercase, generic) get different treatment.
  • Re-transcription needs re-uploading. Adding a term doesn't retroactively fix existing transcripts. Re-process the file to apply the new vocabulary.

Limits​

  • Custom vocabulary is currently English-only. Adding non-English terms is allowed but accuracy improvements vary by language.
  • Very rare terms (made-up brand names) work best. Common-sounding terms may not always override the model's default guess.
  • The list is workspace-wide. If your research domains use very different vocabulary, separate workspaces help.