What is custom vocabulary in dictation?
Why it matters more than headline accuracy
General speech models are trained on common language, so their errors concentrate exactly where your writing is most specific: colleague names, product names, technical terms. A system at 97% overall can still miss your company’s name every single time. Custom vocabulary converts those repeated errors into permanent correct spellings.
How modern tools implement it
Entries can bias recognition toward the term, fix its capitalization and spelling, and sometimes map spoken shorthand to expanded text. Better tools also learn from your corrections, promoting repeatedly fixed words into the dictionary with little manual work.
Building yours quickly
Seed it with the proper nouns of your work: team names, clients, products, tools. Add terms the first time you correct them rather than batching later. In multilingual work, keep the terms of both languages present, so a Japanese sentence can carry an English product name cleanly, and vice versa.
Frequently asked questions
How many terms should I add?
Start with the ten to twenty proper nouns you type most, then grow it organically from real corrections. A short, accurate list beats a huge speculative one.
Does custom vocabulary work across languages?
In tools with multilingual support, entries typically apply regardless of the sentence’s language, which is exactly what mixed-language dictation, such as English product names inside Japanese text, requires.
Keep exploring
Hear it in practice
VoiceOS puts dictation, editing, and voice-to-action in one layer on Mac and Windows.
Try VoiceOS