Using Custom Pronunciation in Outset

Last updated: October 12, 2026

Custom Pronunciation lets you tailor how Outset’s AI moderator says names, brands, and specialized terms. You can also add spelling variants to help recognize those terms in participants’ spoken responses, and provide a definition for context.

This is useful when your study includes a brand name with an unusual pronunciation, an acronym, or a technical term that is easily misheard.


How custom pronunciation works

  • Pronunciation: Controls how the moderator says a word or phrase aloud. For example, you could set “Nike” to “Nai-kee.” In standard interviews, pronunciation changes affect audio only; transcripts and other written content keep the original spelling. In voice-to-voice interviews, the sound-based spelling may also appear in the transcript.

  • Also heard as: Maps transcription variants back to your intended spelling. For example, “Nake” or “Nikke” could be mapped to “Nike.”

  • Definition: Describes what the term means. It helps generate pronunciation suggestions and variants, and provides context for the AI moderator in standard interviews.

Please note: Custom pronunciations apply to all studies in your organization. An entry added or edited from one study is shared across the organization.


Add a word or phrase

You need to be an organization admin with edit access to add, edit, or remove entries. Other active organization members can view entries and listen to pronunciation previews.

  1. Go to Settings → Interview Toolkit → Pronunciations. You can also open Study settings → Pronunciations from the Create page of a study where the moderator speaks. Both places open the same organization-wide list.

  2. Select New.

  3. Enter the Word or phrase using the spelling you want to appear in the interview text.

  4. Add an optional Definition. Keep it short and specific—for example, “Acme is the brand of project management software being discussed in this study.”

  5. Choose the relevant language or languages in the Language column. Use Add another language to give the same term a different pronunciation or set of variants in another language.

  6. Enter a pronunciation using a spelling that reflects how the word sounds. You can also use Suggest pronunciations and select an option. Use the play button to preview it in each selected language before saving.

  7. In Also heard as, enter likely transcription variants, separated by commas, or use Suggest variants. Review suggestions before saving.

  8. Select Save.


Edit or remove an entry

Open Pronunciations and use the search field to find the term. Open the entry’s row actions menu and select Edit to update it, or Remove to delete it. Select Save after editing. To remove an entry, confirm with Delete; this removes the word’s pronunciations, definition, and “Also heard as” variants across its languages.

Saved pronunciation changes apply to interviews that start afterward; an interview already in progress keeps the pronunciation settings it started with.

Remember that these changes affect the shared list across your organization. Check with teammates before changing a term used in other studies.


Tips before you launch

  • Test in each study language. Check how the moderator says the term and how spoken responses are transcribed. Entries apply only to the exact language selected, including regional variants. A pronunciation that works in one language may need a different spelling in another.

  • Choose variants carefully. Variant handling replaces matching words with your intended term. Avoid adding common words that participants may use with a different meaning.

  • Keep definitions concise. Put the most useful meaning first. The moderator receives a shortened version of longer definitions.

  • Keep the original spelling in your guide. Put the sound-based spelling in the pronunciation entry rather than rewriting the term in every question.

  • Check the full interview flow. A pronunciation preview helps you hear the term on its own; a test interview helps you review it in context.


Hope this helps! If you have any further questions, please reach out to our team at support@outset.ai or via chat.