Custom Vocabulary
What is custom vocabulary?
MediaCopilot's custom vocabulary allows you to teach the platform words, names, brands, or technical terms that may be difficult to recognize during automatic audio and video transcription.
When the transcription engine finds a word that sounds similar to one of the variations you've defined, it will automatically replace it with the correct form.
Audio/Video → Transcription → Dictionary applied → Corrected text
Dictionary types
MediaCopilot offers three types of dictionaries, each with a different purpose:
| Type | Purpose | Example |
|---|---|---|
| Domain Dictionary | Corrects poorly transcribed words by replacing them with the correct form. Ideal for proper names, brands, and technical jargon. | Word: "R&D" → Variations: RAID, R and D, rnd |
| Blacklist | Filters and removes unwanted words from the transcription. Useful for censoring inappropriate or irrelevant content. | Profanity, insults, repetitive filler words |
| Whitelist | Prioritizes recognition of certain words so the transcription engine detects them with greater accuracy. | Product names, industry-specific terminology |
Create a dictionary
Go to Customization
Click on your profile icon (top-right corner) and select "Customization".
Profile → Customization → Custom vocabulary
Click "Add new dictionary"
On the Custom vocabulary page, click the "+ Add new dictionary" button and choose the type of dictionary you want to create:
- Domain specific dictionary
- Blacklist
- Whitelist
Give the dictionary a name
Write a descriptive name that helps you identify the dictionary's purpose. For example: "Pharmaceutical terminology", "Team names", "Commercial brands".
Save the dictionary
Once you've added your words (see next section), click "Submit" to save the dictionary.
Add words to the dictionary
Each dictionary entry has two fields:
- Word: The correct form in which you want the term to appear in the transcription.
- Variations: Alternative, incorrect, or phonetic forms that the system might transcribe. Separated by commas.
| Word | Variations |
|---|---|
| I+D | IMASD, I mas D, imaxd |
| Qlik | clic, Clic |
| AC/DC | ACDC |
Apply the dictionary to your videos
Dictionaries are applied through Templates (processing templates). A template defines what actions are executed when processing a video, and among those actions you can include your dictionaries.
Go to Settings → Templates
From your profile menu, go to Settings and then to the Templates tab.
Profile → Settings → Templates
Create or edit a template
Click "+ Add template" to create a new one, or the edit icon next to an existing template.
Enable and assign dictionaries
In the Transcriptions section of the template, check the box for the dictionary type you want to use and click "+ Add dictionary" to select the dictionaries you want to apply.
Upload a video using the template
When uploading a new video from Library → + Upload assets, select the template that contains your dictionary in the "Template" field. You can click "View details" to confirm the dictionary is included.
Tips and best practices
Use cases
- News & Media — Correct names of politicians, parties, cities, and organizations that the system might transcribe incorrectly, especially in regional or minority languages.
- Sports — Create dictionaries with player names, team names, and sport-specific terminology. Ideal for broadcasts where foreign names are common.
- Health & Pharma — Include drug names, active ingredients, and medical terminology for transcriptions of conferences, seminars, or training content in the health sector.
- Technology — Add tool names, frameworks, programming languages, and technical acronyms (e.g., "Kubernetes", "PostgreSQL", "R&D") that are commonly mistranscribed.
- Children's or sensitive content — Use Blacklists to automatically filter profanity or inappropriate content in transcriptions intended for sensitive audiences.
- Regional languages — Improve transcription in co-official or regional languages (Galician, Catalan, Basque, etc.) where local names of people and places are often misinterpreted.