How many terms can I add?

Up to 500 terms per language. RecordX supports a separate vocabulary list for each of the 20 languages the transcription engine covers, so a bilingual team can bias English and Hebrew independently.

What kinds of terms should I add?

Product names, company names, colleagues' full names, technical jargon, medical or legal terminology, drug names, brand names, and any acronyms that come up often. Anything the generic model would not have seen enough of to spell correctly.

Open Settings, pick the language, and paste one term per line into the Custom Vocabulary box. Save. The next transcription in that language uses the updated list.

Does it help with Hebrew names?

Yes. Custom vocabulary is especially useful for Hebrew - given names, company names, and Hebrew technical terms that the generic model gets close but not exact. Add them once and every subsequent Hebrew transcription picks the correct spelling.

Custom vocabulary - teach RecordX your names and terms

Paste up to 500 terms per language into Settings and RecordX biases every subsequent transcription toward that spelling. Live today, one list per language, twenty languages supported.

How it works

Open Settings, pick a language, paste one term per line into the Custom Vocabulary box. Save. The transcription engine picks up the list on the next job and biases word choice toward those spellings when it hears a close phonetic match. Each language keeps its own list, so a bilingual team can bias English one way and Hebrew another without cross-contamination.

What to add

Company and product names - your product, competitors' products, partner brands. The generic model has never heard of them; a vocabulary term fixes that in one line.
Colleagues' full names - given name plus surname, especially names with unusual spellings.
Technical jargon and acronyms - anything specific to your domain that generic training data underweights.
Medical, legal, or scientific terminology - drug names, procedure names, statute citations, species names. The engine gets close but rarely exact without hints.

Use cases

Sales calls that mention competitor names - a Salesforce vs. HubSpot vs. Pipedrive conversation is only useful if the transcript spells them correctly.
Medical dictation with drug names - Ozempic, dupilumab, methylphenidate all sound generic to a base model. A short vocabulary list fixes them all.
Legal review with case citations - party names and statute references stay accurate for downstream search.
Hebrew names and terms - Hebrew given names and Hebrew technical terminology benefit most, since the generic model has less coverage per name than English.

Works across the twenty languages

Custom vocabulary is available in every language RecordX transcribes - Hebrew, English, Spanish, French, German, Portuguese, Italian, Russian, Ukrainian, Arabic, Turkish, Chinese, Japanese, Korean, Hindi, Indonesian, Vietnamese, Thai, Romanian, and Czech. See the multi-language page for the full list and the transcription feature for the underlying pipeline. The list is applied to every capture path - Zoom, Google Meet, Microsoft Teams, the Chrome extension, the mobile app, and the softphone.

Start for free