TrueConf AI Server 1.1: Detecting Speakers by Voice, Uploading Recordings, and Transcription Search

We are pleased to present TrueConf AI Server 1.1, the latest update of our conference transcription server. This release features the ability to transcribe your audio and video recordings with voice separation, expanded transcription search capabilities, improved work with summaries, and increased system stability.

Transcription of voice-separated recordings

TrueConf AI Server 1.1 introduces a diarization module that identifies individual speakers by voices in uploaded audio and video recordings. These can be either conference recordings or third-party files. The system automatically detects multiple speakers, performs speech recognition for each, and compiles a single transcript indicating the speakers.

After processing the recording, a user can manually assign a name to each recognized speaker.

Searching transcriptions

We’ve added a global search for all transcripts available to a user in the AI server personal area, as well as in the corresponding section of TrueConf client applications. Entering a word into the search bar displays a list of all the transcripts with that text which are available to the user.

When you open the selected transcript, the search for the entered text will continue.

Support for Markdown formatting in summaries

Conference summaries now support Markdown syntax, allowing you to format the text using bold, italics, headings, lists, and tables.

Two audio playback modes

When viewing a transcript in your personal area, two modes of the built-in player are available:

• Auto-follow mode. ПThe player automatically scrolls to the active phrase as the audio plays.
• Manual mode. It is activated when scrolling or navigating to a specific location via search. In this case, the player does not follow the lines, but remains at the current position.



Other changes and improvements


• We’ve improved the quality of final summaries thanks to the new AI model
• We’ve improved support for working on the server with encryption enabled
• We’ve improved server stability when using multiple graphics cards
• The “Partially transcribed” status is now displayed next to the conference recording, which was not fully transcribed
• The “License information” section of the control panel now displays the versions of each TrueConf AI Server component
• When updating the AI server, the console displays a list of configuration files that do not match the current release version
• The maximum size of a text for testing a prompt is limited to 100,000 characters
• The button for starting summarization is now hidden for the users with read-only access
• The license agreement has been moved from the registration page to the administrator login page
• The overall stability of the transcription server has been improved.

You can view the full list of changes in our changelog.