There is one more tool, SoundJack (
www.soundjack.eu).
The problem with any of these tools is that to keep the latency small, UDP is used for transport which means that packets can get lost. If too many packets get lost or at a large amplitude you will get dropouts and more or less loud impulses.
Some weeks ago there was a video at the starting page of SoundJack (no longer there, maybe in a different subpage) with two musicians (I think both living in Vienna), a saxophone player and a accordion player who did an amazing video.
But they used tricks to make it. They used SoundJack to play synchronized (using a Focusrite Scarlett Interface), but at the same time recorded the microphone sound on each end with a second computer (it is not a good idea to do both on one machine as you get more dropouts). Then they transferred one of the recordings to the other person to combine both.
There is still the problem of different sample clocks during the both recordings, so the length (and pitch) needs to be adjusted to be synced during the whole piece. They also recorded video, so they are both visible during their original playing.
Latency is a big problem. The computer has latency due to other processes interrupting, routers and repeaters have latency, packets might get delayed or rerouted due to high packet traffic.
The internal machine latency could be dropped to 5 to 6 ms, the route to the mirror in Nürnberg (which is about 400 km from me) takes about 24 to 26 ms (reducing the number of packet buffers to two in both cases).
The guys from Vienna were in the same town and used the same internet provider and got a latency from a bit more than 10 ms.
Macintosh computers have a bit less latency (with a CoreAudio driver) than Windows machines with an ASIO driver.
Also the buffer size is important. A small buffer size helps reducing latency, but puts a lot more load on the machine. That can lead to dropped packets too and it depends on the interface and driver how low you can go.
number of participants is limited to about 6 other players or a mix down server (adding more latency) is needed. Also a more or less fast internet connection for all participants (in and out!).
A synchronized metronome is planned for SoundJack, but not realized yet.
To use the standard connection with SoundJack, a router needs to forward UDP port 50050 to the IP address of the SoundJack machine. There are hints on the website how to do that.
This is not needed for the mix down server, but that adds delay. (BTW, each participant can make his SoundJack to a mix down server, no need to use a distant public one.)
Video is possible but better not used (experimental, high traffic or bad image quality).
The Focusrite (Solo is enough, maybe add an impedance buffer if using a piezo pickup) are the best small interface with a good quality, the Behringer UR-series is a cheap alternative with low latency, but with a less good mic audio quality (but instrument and line input might be OK).
My MOTU FireWire interfaces work as well as the Focusrite. But on a Windows machine you need a good FireWire interface (TI or Ricoh, not VIA or NEC) and even then it doesn’t work as well as on a Macintosh. For Windows the Focusrite works a bit better.
My Zoom B2.1 works as an interface too, but has the original sound on the headphones too. On the Mac I can use different input and output (not possible with the ASIO Windows driver), so I can route the output to the internal Mac headphones, but to talk I have to switch between Zoom and internal mic which is no fun with the bass.
I have heard (but not tested myself) that the latency of a Zoom field recorder is too large to use it as an interface for live music.
So things are possible, but technically demanding.
You can try SoundJack with internal audio first. Works well on the Mac and not too bad on a Windows machine with ASIO4ALL. No need to buy anything. Sing or clap (or play) to find out if sync works. If you use a mic you need headphones to avoid feedback.
The author of SoundJack is currently on vacation. But there is a chat window in SoundJack and users try to help.
Due to Alexander Carot, the author, who wrote his doctoral dissertation about real-time network audio collaboration, the maximum distance is 700 km. But time delay by routing in the network transfer can reduce that a lot.