This document explains how to configure custom sensitive words by using the built-in Speech-to-Text (STT) node.
Overview
The custom sensitive words feature lets you define a word list. The system then recognizes words from this list in real time and redacts them in client-side subtitles by replacing them with an asterisk (*).
After a word is redacted, the system passes the redacted content to the large language model (LLM) node.
Procedure
-
Prepare a sensitive word file: Create a plain text (.txt) file with the sensitive words you want to block.
Note-
The language of your sensitive words must match the language model configured in the console.
-
Each file can contain a maximum of 500 words. Each word can be up to 10 characters long. The file size cannot exceed 100 KB.
-
The file must be saved in
UTF-8encoding. The sensitive words must not contain punctuation or special characters.
-
-
Configure the STT node: Upload your sensitive word file to the Speech-to-Text (STT) node in your real-time workflow.
In the Basic information section of the STT node's configuration panel, find the Custom sensitive words field and click Upload.
-
Apply the changes: After you upload the file, save or create the workflow to apply the new configuration.