Your English is amazing, if you wouldn't have said anything I wouldn't
have noticed you are not a native speaker.

Although I don't have any experience in this field I can offer my
creative input.

Since at a talk content comes before quality, you could simply use two
surveillance webcams with Ethernet ports (they usually include
microphones, some have jacks for external microphones). They give you a
finished audio/visual stream in a format you can directly relay to the
Internet.

Then all you need is a PC with a streaming server that will allow you to
select which of the incoming streams is to be relayed to the outgoing
stream. You can relay the outgoing stream LIVE to the internet as well
as recording it.

Many of these cams stream MPG-2 which is supported by most platforms.

Regardless of whether or not you use this specific solution, I would in
any case let the CAMERAS rather than the computer synchronize audio and
video. Letting the cameras handle the web-encoding is an added benefit.

If you really want to encode on the computer, you might look into the
free and open source combination of Theora for video and Vorbis for audio.

Carlo