 | In today’s hybrid work environment, clear communication and effective collaboration are more important than ever. Microsoft Teams Rooms on Windows devices now feature AI-powered speaker recognition technology, expanding its capabilities to all microphones, not just intelligent speaker equipment. This is a huge step in bringing Copilot AI functionality into Teams Rooms. In the past, Microsoft required specific Intelligent Speaker hardware to enable this functionality, but this is no longer necessary as any MTR will be able to provide this capability. In this blogpost, we take a first look at this newly released feature for Teams Rooms on Windows and experience how the new cloud-powered AI speaker recognition makes a difference in an actual meeting. |
What is Speaker Recognition?
Speaker recognition technology analyzes the unique vocal characteristics of each participant, creating a “voiceprint” similar to a fingerprint. This allows Teams Rooms to identify and attribute each speaker accurately during meetings which in turn allows Copilot to generate responses to chat questions on what each participant in the room has spoken. Without speaker recognition and attribution in the transcript, Copilot can only respond to queries on what someone in the room has said, and not who specifically has said it.
Benefits of Speaker Recognition
How to Enable Speaker Recognition
Firstly, users who have a Copilot license can create a voice profile in the Teams desktop app. But first IT admins need to enable this feature for their organization using Teams meeting policies and PowerShell scripts. The powershell cmdlet to enable this at the global level for all users is:
set-CsTeamsMeetingPolicy -EnrollUserOverride Enabled -identity Global -RoomAttributeUserOverride Attribute -RoomPeopleNameUserOverride On -AllowTranscription $True
The output of this is shown below:
Speaker recognition technology analyzes the unique vocal characteristics of each participant, creating a “voiceprint” similar to a fingerprint. This allows Teams Rooms to identify and attribute each speaker accurately during meetings which in turn allows Copilot to generate responses to chat questions on what each participant in the room has spoken. Without speaker recognition and attribution in the transcript, Copilot can only respond to queries on what someone in the room has said, and not who specifically has said it.
Benefits of Speaker Recognition
- Noise Suppression: Background noise is minimized, making conversations clearer.
- Intelligent Recap: Copilot can provide detailed summaries and insights by identifying who said what.
- Enhanced Transcripts: Meeting transcripts are more precise, attributing comments to the correct individuals.
How to Enable Speaker Recognition
Firstly, users who have a Copilot license can create a voice profile in the Teams desktop app. But first IT admins need to enable this feature for their organization using Teams meeting policies and PowerShell scripts. The powershell cmdlet to enable this at the global level for all users is:
set-CsTeamsMeetingPolicy -EnrollUserOverride Enabled -identity Global -RoomAttributeUserOverride Attribute -RoomPeopleNameUserOverride On -AllowTranscription $True
The output of this is shown below:
Once this is done, users can now access the Recognition tab in their Teams client and record their voice profile as well as their facial profile which is then securely stored in the Teams cloud. This is shown in the screenshot below:
Users can update or remove their voice and face profiles at any time. Now that we've created the profiles, we can proceed to join a meeting from an MTR on Windows device and see if the new AI powered speaker recognition can correctly identify the speaker in the room. Below is a recap of the short 5min meeting where I am joining from my HP G9 MTR PC connected to a HP Poly Studio R30 speaker bar: Note that it already has identified me correctly from my voice as can be seen in the timelines below the video as I playback the meeting:
It's worth nothing that there's a small <1% of the transcript where it did not recognize my voice and showed "Speaker 1" instead, I've attached the full transcript of the meeting at the bottom of this blogpost for anyone who wishes to see the details.
Security and Privacy
Microsoft ensures that voice and face data are encrypted and stored securely, complying with GDPR and other privacy regulations6. Users can unenroll their profiles at any time, with data automatically deleted if not used for a year.
Future Developments
This exciting new feature will soon be available for Teams desktop in BYOD meeting spaces and Teams Rooms on Android devices later this year. While hardware based intelligent speakers offer the highest quality, this update makes advanced speaker recognition accessible to more users on any MTR device, without any special hardware.
Microsoft ensures that voice and face data are encrypted and stored securely, complying with GDPR and other privacy regulations6. Users can unenroll their profiles at any time, with data automatically deleted if not used for a year.
Future Developments
This exciting new feature will soon be available for Teams desktop in BYOD meeting spaces and Teams Rooms on Android devices later this year. While hardware based intelligent speakers offer the highest quality, this update makes advanced speaker recognition accessible to more users on any MTR device, without any special hardware.
Full Transcript
0:0:0.0 --> 0:0:20.760
Bren Kwok
OK. Good afternoon, everyone. Thanks for joining this session on the AI innovations that are transforming our world. Ladies and gentlemen, today I'd like to share with you some of the groundbreaking advancements in artificial intelligence. They have reshaped our lives in recent times.
0:0:22.200 --> 0:0:35.880
Bren Kwok
Mostly the revolution of generative AI tools. Generative AI tools like ChatGPT or Copa, for example, have achieved mass adoption at an unprecedented rate.
0:0:37.0 --> 0:0:40.560
Bren Kwok
These neural networks can generate human like text.
0:0:41.880 --> 0:0:45.720
Bren Kwok
Revolutionising content creation customer service.
0:0:47.310 --> 0:0:54.310
Bren Kwok
Imagine a world where AI can draught reports, summarise emails and even create.
0:0:54.490 --> 0:1:13.970
Bren Kwok
Entire slide decks in seconds. Microsoft and Google have made this a reality. Putting chat botbased assistance into the hands of billions of users. Secondly, image making models from pixels to art.
0:1:15.540 --> 0:1:23.700
Bren Kwok
Meta introduced image making models that allow users to generate shareable images effortlessly.
0:1:25.20 --> 0:1:34.60
Bren Kwok
With a single click you can create visual mashups turning Mickey Mouse and SpongeBob SquarePants into pilots flying into the sky.
0:1:35.370 --> 0:1:45.90
Bren Kwok
Google's new phones use AI to enhance photos, transforming sad faces into happy ones and dull afternoons into perfect sunsets.
0:1:47.150 --> 0:1:49.350
Bren Kwok
Thirdly, open source AI and licencing.
0:1:50.990 --> 0:1:57.670
Speaker 1 (15fd1e24-ad70-47e4-8ee2-264b998173fe@d4deddde-b799-4bb1-ac61-ac5aeebb2499)
The AI community has witnessed a significant development in open source AI.
0:1:58.110 --> 0:2:4.350
Bren Kwok
Researchers and practitioners collaborated, sharing knowledge and code to accelerate progress.
0:1:50.790 --> 0:2:4.470
Bren Kwok
The AI community has witnessed a significant development in open source AI. Researchers and practitioners collaborated, sharing knowledge and code to accelerate progress.
0:2:5.790 --> 0:2:16.870
Bren Kwok
Licencing debates around AI models raised critical questions about ownership, access and ethical implications. Last but not least.
0:2:18.270 --> 0:2:20.230
Bren Kwok
AI advancements across sub areas.
0:2:21.120 --> 0:2:25.120
Bren Kwok
In the last five years, AI has made remarkable progress.
0:2:27.730 --> 0:2:28.850
Bren Kwok
Various domains.
0:2:30.130 --> 0:2:53.850
Bren Kwok
For example, in vision there is now improved image recognition and generation in natural language processing. There's enhanced understanding and generation of text in robotics. The integration of vision and motor control into a humanite robot. Multi Agent systems such as coordinated decision making systems.
0:2:55.100 --> 0:3:7.660
Bren Kwok
Speech recognition and generation have become more accurate and expressive, and finally image and video generation have now given us stunning visual content.
0:3:8.940 --> 0:3:18.860
Bren Kwok
In conclusion, we stand at the cusp of an AI driven future as we marvel at these innovations, let us also grapple with the impact.
0:3:20.460 --> 0:3:24.540
Bren Kwok
What's looking back at us from the A from the AI abyss, only time will tell.
0:3:25.600 --> 0:3:33.520
Bren Kwok
Thank you for your attention and may our journey into the AI frontier be both exciting and thoughtful. Thank you.
0:0:0.0 --> 0:0:20.760
Bren Kwok
OK. Good afternoon, everyone. Thanks for joining this session on the AI innovations that are transforming our world. Ladies and gentlemen, today I'd like to share with you some of the groundbreaking advancements in artificial intelligence. They have reshaped our lives in recent times.
0:0:22.200 --> 0:0:35.880
Bren Kwok
Mostly the revolution of generative AI tools. Generative AI tools like ChatGPT or Copa, for example, have achieved mass adoption at an unprecedented rate.
0:0:37.0 --> 0:0:40.560
Bren Kwok
These neural networks can generate human like text.
0:0:41.880 --> 0:0:45.720
Bren Kwok
Revolutionising content creation customer service.
0:0:47.310 --> 0:0:54.310
Bren Kwok
Imagine a world where AI can draught reports, summarise emails and even create.
0:0:54.490 --> 0:1:13.970
Bren Kwok
Entire slide decks in seconds. Microsoft and Google have made this a reality. Putting chat botbased assistance into the hands of billions of users. Secondly, image making models from pixels to art.
0:1:15.540 --> 0:1:23.700
Bren Kwok
Meta introduced image making models that allow users to generate shareable images effortlessly.
0:1:25.20 --> 0:1:34.60
Bren Kwok
With a single click you can create visual mashups turning Mickey Mouse and SpongeBob SquarePants into pilots flying into the sky.
0:1:35.370 --> 0:1:45.90
Bren Kwok
Google's new phones use AI to enhance photos, transforming sad faces into happy ones and dull afternoons into perfect sunsets.
0:1:47.150 --> 0:1:49.350
Bren Kwok
Thirdly, open source AI and licencing.
0:1:50.990 --> 0:1:57.670
Speaker 1 (15fd1e24-ad70-47e4-8ee2-264b998173fe@d4deddde-b799-4bb1-ac61-ac5aeebb2499)
The AI community has witnessed a significant development in open source AI.
0:1:58.110 --> 0:2:4.350
Bren Kwok
Researchers and practitioners collaborated, sharing knowledge and code to accelerate progress.
0:1:50.790 --> 0:2:4.470
Bren Kwok
The AI community has witnessed a significant development in open source AI. Researchers and practitioners collaborated, sharing knowledge and code to accelerate progress.
0:2:5.790 --> 0:2:16.870
Bren Kwok
Licencing debates around AI models raised critical questions about ownership, access and ethical implications. Last but not least.
0:2:18.270 --> 0:2:20.230
Bren Kwok
AI advancements across sub areas.
0:2:21.120 --> 0:2:25.120
Bren Kwok
In the last five years, AI has made remarkable progress.
0:2:27.730 --> 0:2:28.850
Bren Kwok
Various domains.
0:2:30.130 --> 0:2:53.850
Bren Kwok
For example, in vision there is now improved image recognition and generation in natural language processing. There's enhanced understanding and generation of text in robotics. The integration of vision and motor control into a humanite robot. Multi Agent systems such as coordinated decision making systems.
0:2:55.100 --> 0:3:7.660
Bren Kwok
Speech recognition and generation have become more accurate and expressive, and finally image and video generation have now given us stunning visual content.
0:3:8.940 --> 0:3:18.860
Bren Kwok
In conclusion, we stand at the cusp of an AI driven future as we marvel at these innovations, let us also grapple with the impact.
0:3:20.460 --> 0:3:24.540
Bren Kwok
What's looking back at us from the A from the AI abyss, only time will tell.
0:3:25.600 --> 0:3:33.520
Bren Kwok
Thank you for your attention and may our journey into the AI frontier be both exciting and thoughtful. Thank you.
RSS Feed