VIDEO CONFERENCE TRANSCRIPT QUERYING USING ARTIFICIAL INTELLIGENCE
Techniques for video conference transcript querying using artificial intelligence are provided. In an example method, a video conference provider joins a client device to a video conference. The video conference provider receives a deletion election. The video conference provider receives an audio stream from the client device and generates a portion of a transcript of the video conference. Prior to the video conference concluding, the video conference provider processes the portion of the transcript to configure a large language model (“LLM”) to respond to queries based on the video conference. The video conference provider receives a query relating to the video conference and causes the LLM to process the query and the portion of the transcript. The video conference provider outputs a response, generated by the LLM, to the query. The video conference provider then deletes the portion of the transcript based on the deletion election.
1 . A method, comprising:
joining a first client device of a plurality of client devices to a video conference hosted by a video conference provider;
receiving a deletion election;
receiving an audio stream from the first client device;
generating, based on the audio stream, a portion of a transcript of the video conference;
prior to the video conference concluding:
processing the portion of the transcript to configure a large language model (“LLM”) to respond to queries based on the video conference, including the portion of the transcript;
receiving a query relating to the portion of the transcript;
causing the LLM to process the query and the portion of the transcript; and
outputting a response, generated by the LLM, to the query; and
deleting the portion of the transcript based on the deletion election.
2 . The method of claim 1 , wherein deleting the portion of the transcript is responsive to the video conference concluding.
3 . The method of claim 1 , wherein the transcript comprises a plurality of portions.
4 . The method of claim 3 , wherein the transcript is generated at the conclusion of the video conference.
5 . The method of claim 4 , further comprising:
generating the transcript, comprising:
generating a plurality of portions of the transcript, comprising:
determining, from the audio stream, one or more audio stream portions;
responsive to an end of meeting signal, generating the plurality of portions of the transcript using the one or more audio stream portions; and
deleting the audio stream and the one or more audio stream portions.
6 . The method of claim 1 , wherein the transcript is continuously generated in near-real-time during the video conference comprising continuously generating a plurality of portions of the transcript.
7 . The method of claim 6 , wherein generating the portion of the transcript of the video conference comprises:
determining, from the audio stream, an audio stream portion; and
generating, based the audio stream portion, the portion of the transcript.
8 . The method of claim 6 , wherein generating the portion of the transcript of the video conference comprises:
determining, from the audio stream, one or more first audio stream portions;
generating, based the one or more first audio stream portions, a first portion of the transcript;
determining, from the audio stream, one or more second audio stream portions;
generating, based the one or more second audio stream portions, a second portion of the transcript; and
generating the portion of the transcript comprising combining the first portion of the transcript and the second portion of the transcript.
9 . The method of claim 8 , wherein the portion of the transcript includes the transcript of the video conference up through the receiving the query.
10 . A non-transitory computer-readable storage medium storing processor-executable instructions configured to cause one or more processors to:
join a first client device of a plurality of client devices to a video conference hosted by a video conference provider;
receive a deletion election;
receive an audio stream from the first client device;
generate, based on the audio stream, a portion of a transcript of the video conference;
prior to the video conference concluding:
process the portion of the transcript to configure a large language model (“LLM”) to respond to queries based on the video conference, including the portion of the transcript;
receive a query relating to the portion of the transcript;
cause the LLM to process the query and the portion of the transcript; and
output a response, generated by the LLM, to the query; and
delete the portion of the transcript based on the deletion election.
11 . The non-transitory computer-readable storage medium of claim 10 , wherein the instruction to delete the portion of the transcript is responsive to the video conference concluding.
12 . The non-transitory computer-readable storage medium of claim 10 , wherein the transcript comprises a plurality of portions.
13 . The non-transitory computer-readable storage medium of claim 12 , wherein the transcript is generated after the conclusion of the video conference.
14 . The non-transitory computer-readable storage medium of claim 10 , wherein the transcript is continuously generated in near-real-time during the video conference comprising continuously generating a plurality of portions of the transcript.
15 . The non-transitory computer-readable storage medium of claim 14 , wherein:
the instruction to generate the portion of the transcript of the video conference comprises:
determining, from the audio stream, one or more first audio stream portions;
generating, based the one or more first audio stream portions, a first portion of the transcript;
determining, from the audio stream, one or more second audio stream portions;
generating, based the one or more second audio stream portions, a second portion of the transcript; and
generating the portion of the transcript comprising combining the first portion of the transcript and the second portion of the transcript; and
the portion of the transcript includes the transcript of the video conference up through the receiving the query.
16 . A system comprising:
one or more non-transitory computer-readable media; and
one or more processors communicatively coupled to the one or more non-transitory computer-readable media, the one or more processors configured to execute processor-executable instructions stored in the non-transitory computer-readable media to:
join a first client device of a plurality of client devices to a video conference hosted by a video conference provider;
receive a deletion election;
receive an audio stream from the first client device;
generate, based on the audio stream, a portion of a transcript of the video conference; <prior to the video conference concluding:
process the portion of the transcript to configure a large language model (“LLM”) to respond to queries based on the video conference, including the portion of the transcript;
receive a query relating to the portion of the transcript;
cause the LLM to process the query and the portion of the transcript; and
output a response, generated by the LLM, to the query; and
delete the portion of the transcript based on the deletion election.
17 . The system of claim 16 , wherein the instruction to delete the portion of the transcript is responsive to the video conference concluding.
18 . The system of claim 16 , wherein the transcript comprises a plurality of portions.
19 . The system of claim 18 , wherein the transcript is generated after the conclusion of the video conference.
20 . The system of claim 16 , wherein:
the transcript is continuously generated in near-real-time during the video conference comprising continuously generating a plurality of portions of the transcript;
the instruction to generate the portion of the transcript of the video conference comprises:
determining, from the audio stream, one or more first audio stream portions;
generating, based the one or more first audio stream portions, a first portion of the transcript;
determining, from the audio stream, one or more second audio stream portions;
generating, based the one or more second audio stream portions, a second portion of the transcript; and
generating the portion of the transcript comprising combining the first portion of the transcript and the second portion of the transcript; and
the portion of the transcript includes the transcript of the video conference up through the receiving the query.