Chapter 9: Lexicons and Voiceprints
Teach Scribis special terms (ASR Dictionary), fix common mistakes automatically (Correction Dictionary), pronounce names correctly (Phonetic Dictionary), and recognize each person's voice (Voiceprint Recognition).
Chapter 9: Lexicons and Voiceprints (Pro)
💡 This chapter covers advanced features. If you're just starting out with Scribis, feel free to skip it and come back when some names keep getting recognized wrong, or you find yourself renaming speakers by hand all the time.
Lexicons & Voiceprints in the left menu is where you teach Scribis your names, terms and voices, so it makes fewer mistakes. It has five sections:
| Section | What it's for | In one sentence |
|---|---|---|
| Voiceprint Recognition | Recognize who is speaking | Subtitles say "Host Amy" instead of "Speaker 1" |
| Phonetic Dictionary | Help AI voices pronounce words correctly | For example, get an AI voice-over to say "Scribis" correctly |
| Correction Dictionary | Fix mistakes automatically | Whenever it sees "Scrivis," change it to "Scribis" |
| ASR Dictionary | Help speech recognition understand special terms | Tell the model in advance which words to expect |
| Prompt Styles | Save AI instructions you use often | For example a "formal written English" writing style |
9.1 ASR Dictionary: teach the model special terms
"Hotwords" are words you want the model to listen out for, such as product names, people's names and technical terms.
- Click ASR Dictionary and stay on the Hot Words tab.
- Click Add Hotword at the top right, choose a language and type the word (for example
ScribisorKubernetes). - When you transcribe, turn on Use Hotwords Dictionary in Advanced Settings (the hotwords dictionary is this list of hotwords).

💡 Only some speech models support hotwords (SenseVoice Small, ElevenLabs Scribe v2 and Gemini Transcribe). If those names don't mean anything to you, skip this for now. With other models, write the special terms in the Prompt box when you transcribe, for example "This recording mentions Scribis."
9.2 Correction Dictionary: fix mistakes automatically
Each rule is "wrong spelling → correct spelling": the wrong spelling goes on the left and the correct spelling on the right (the column names Incorrect Spelling (Key) and Correct Spelling (Value) just mean left and right).
The Optimize & Translate window calls this same set of rules your replacement dictionary. It's the Correction Dictionary tab on this page; the three names all mean the same place.
- Click Correction Dictionary and switch to the Correction Dictionary tab.
- Click Add Entry. For example, enter
slight deckas the incorrect spelling andslide deckas the correct one. - To check that a rule works, type a sentence into the Text Correction Test box below.

These rules are applied automatically when you tick Apply my replacement dictionary first in Optimize & Translate. Corrections you approve while optimizing can also be saved here in one click (see Chapter 6, section 6.4).
⚠️ Rules apply to every place the text appears. Don't use very common words (such as "the" or "is") as an incorrect spelling, or they'll be changed where they shouldn't be.
The Translation Dictionary tab on the same page manages the word pairs used when translating.
9.3 Phonetic Dictionary: help AI voices say it right
When you use speech synthesis (AI voice-over), some abbreviations or names may be mispronounced. In Phonetic Dictionary, click Add Entry, type the word, choose the language and write how it should be pronounced (for example, type Scribis and write how it sounds). You don't need to know the International Phonetic Alphabet; just follow the hints on screen.

9.4 Voiceprint Recognition: remember each person's voice
Useful for regular meetings, podcasts or interview series with the same people.
![]()
First-time setup:
- Turn on Speaker Diarization when you transcribe (Chapter 3, section 3.4).
- In the subtitle editor, click Speaker Settings at the top and rename "Speaker 1", "Speaker 2" and so on to people's real names.
- Named speakers appear in Voiceprint Recognition marked Ungrouped.
- Click Manage Groups to create a group (for example "Weekly product meeting"), then use Assign group on each row to put the speakers in it.
After that:
The next time you transcribe the same group of people, Scribis compares their voices and labels them by name, so you don't have to fix the names by hand every time.
⚠️ A voiceprint is a pattern made from someone's voice, a bit like a fingerprint. Before creating a voiceprint for someone else, tell them and get their permission.
You can also click Add Record, enter a name and choose a clear recording of that person (an audio file) to create a voiceprint directly.
9.5 Prompt Styles
In Prompt Styles, click Add Style to save AI instructions you use often, for example "Summarize in bullet points using plain English." You can then pick the style when using AI features instead of typing it again each time.