These are rough notes.

tl;dr

• If you are unsure, Zoom is the de facto standard, so start with Zoom.
• If you can clarify requirements, consider other services too.
• If you use G Suite, Google Meet is good and worth trying.

Requirements to consider

• How much security do you need?
• How easily should people be able to join?
• How stable does communication need to be? How many participants should you expect?
• Should you assume participation from mobile or tablet apps?
• Will the same team use it regularly, or will there constantly be first-time users?
• Do you want to record?

Summary of various video conferencing systems

• Zoom: becoming the de facto standard. It can be used for free. We also use it for regular meetings in our technology team. It is the most stable and has many features. Conversely, there is more friction, such as installing the application and registering an account.
• Whereby, formerly Appear: can be used by simply sharing a URL. It can also be used for free. It is easy, but security is weaker for that reason. Also, speed tends to be unstable.
• Google Meet: can be used for free if you use G Suite. You can join by sharing a URL, and because there is approval before entering the room, a certain level of security is also ensured. The connection is stable, and around 30 participants is not a problem.

Not video conferencing, but seminar streaming

• YouTube Live: at present, it is almost the only choice. Zoom also has webinar functionality on paid plans, so if you already pay for Zoom, it can be an option.

Tools that can do video conferencing as an extension of chat tools, unverified

• Slack
• Discord
• Microsoft Meet

Messaging tools with video support, unverified

• Skype
• Facebook Messenger
• Google Duo or Google Hangout. Note that Google Meet is separate.

Unusual options

• https://pragli.com/ meetings with avatars
• https://remo.co/conference/ join by sitting in rooms inside a virtual space
• https://tandem.chat/ intended for pair programming
• https://jitsi.org/jitsi-meet/ open source and can also run on your own server

Build-it-yourself options

• Full scratch implementation
• https://www.twilio.com/
• https://webrtc.ecl.ntt.com/ NTT-related

Technical notes, because understanding the technology makes service differences easier to understand

• Browser-based video conferencing basically uses WebRTC.

→ Software installation gives more freedom, so optimization is easier.

→ → In other words, convenience without installation is a trade-off against comfortable communication and richer functionality.

→ → → That said, even install-based services are still bad if they are not implemented properly.

• On mobile, WebRTC APIs are not fully supported.

→ Apps become necessary.

→ → If it is only live streaming, it is not WebRTC, just watching video, so it can be viewed.

• Video packets are basically sent over UDP.

→ So it is basically a mechanism for sending large amounts of data.

→ → Services with comfortable communication optimize this data.

→ → → Method 1 is installing software and compressing through that software.

→ → → Method 2 is using an SFU, described below, and optimizing through an intermediate server.

• WebRTC mechanisms include P2P and SFU, Selective Forwarding Unit.

→ P2P literally connects devices directly.

→ → As the number of participants increases, the number of connections grows exponentially, so once four or more people join, it becomes difficult.

→ SFU prepares an intermediate server. Clients communicate only with the SFU, so basically it is no different from a one-to-one connection. Strictly speaking, download data increases, so in many cases the number of videos that can be displayed is limited.

→ → Even with SFU, there is room for filtering data to some extent, so I think companies with strong development capabilities implement that optimization.

→ → → A further path is MCU, Multipoint Control Unit, a server approach that connects multiple points and composites them in real time. Twilio is SFU, so it does not work like a quick implementation through an API. Apparently this used to be the de facto approach. Because it composites in real time, it can apparently combine 100 videos into one stream and distribute it.