LiveKit Community

It should be pretty straight forward for what you are looking for. You can selective subscribe to tracks depending on proximity. There are a bunch of spatial/virtual worlds that have integrated/integrating livekit now.

What platform is the game built on?

windows. There are serveral game clients developed with different language. So i plan to use both livekit and game sdk to build this voice client.

Screen Shot 2022-04-18 at 1.07.37 PM.png

funny that you ask. <@U02EU2ZPTQU> just published it on unpkg: <https://unpkg.com/livekit-client>

You can follow the main repos on GitHub. You can either follow all activity (if you want to see PR discussions and posted bugs), or subscribe to specific events like releases. Personally, I follow <https://github.com/livekit/livekit-server> and the client SDK that my project is using (JS).

Thanks for your feedback <@U02C10ZEP8R>. That's pretty much what I'm doing right now. :slightly_smiling_face:

We will be blogging/writing about some of these bigger changes soon!

But keeping a changelog in the repo would also be a good idea.

<@U01KZ61GV5Z> If I had one bit of feedback, it would be that I would love it if there was a little bit more context in the PRs and commits that come through. I imagine a lot of things are coming from internal goals &amp; discussions. Things are moving fast on the project and it'd be great to be able to understand some of the reasons for changes coming in so we can take advantage of and test the new features as they get implemented :-)

<@U02C10ZEP8R> I hear you! It's a great point to make them more digestable as well. Will relay that feedback to the team. 

Kay, this is going to be a significant lift. It'll not be for the faint of heart. P2P does not mesh with the livekit model very well, especially when it has to switch to SFU mode.

Thank you David. Will discuss your inputs with the team and try to make them see the point!!

Even it it weren't difficult to implement, I'd recommend against it. I built a product previously using Jitsi, which has this exact model (P2P for 1:1, SFU for more). I ended up disabling that option in my product because it did not work as well as having the SFU always in use.

The problem, I believe, is network peering. Logically it makes sense that `User &lt;-&gt; User`  would be more efficient than `User &lt;-&gt; SFU &lt;-&gt; User`,  but that assumes that the network path between the users is just as good as the network path between each user and the SFU.

Many ISPs, especially the large ones, have direct peering agreements with large colocation centers. In major cities, it's entirely possible that you are getting onto a major provider's backhaul network without ever leaving the data center that your ISP is in. However, most ISPs don't peer directly with other ISPs and even users on the same ISP may go through some extra hops to get to each other because they are purposefully segregated. Because of this, it is often a better and lower latency connection to go through a server located in a datacenter than it is to directly  connect two clients.

The only benefit that you would get is lower bandwidth costs on your LiveKit server. But that would be at the expense of the user experience.

thanks, await one async method right? which method do I need to await, because I can not find 'the connect', what does it refers to

this one: <https://docs.livekit.io/client-sdk-flutter/livekit_client/Room/connect.html>

I have some question about your benchmarking scenario for large meetings ( as described here <https://docs.livekit.io/deploy/benchmark/|https://docs.livekit.io/deploy/benchmark/>)
You say that you have 100 pubs amd 100 subs in one meeting, everyone is publishing video and audio. It means that we have 100 x 2 = 200 published tracks.
So, my question is about the number of received tracks per participant.
I assume that every participant receives audio and video tracks from all other participants (198 received tracks per participant). Am I right ?

If you're using example app, what's the size of grid for layout ?

Next, if we have 5x5 grid, each participant receives 25 video streams, so 25 x 100 = 2500 streams at all

Unfortubately, I have no resources to do real load testing and simulate estimated number of participants

My goal is to find out an average number of streams per 1 virtual cpu core.

I would appreciate it if you test it yourself and upload the data

are you calling <https://docs.livekit.io/client-sdk-flutter/livekit_client/Room/disconnect.html|Room.disconnect> when the user leaves?

If we configure redis, when one party hangs up,the participant will not receive the notification until 30 seconds later

I remember seeing someone adding a Virtual Background to the video feed in livekit.. so I would say so.

Hey <@U03DCV6VCLT>, it's not possible to process the frames with Flutter (yet). it's on our mind though. with the native SDKs you can already to this by processing the frames and then making a video source out of them.

<@U032TUAL4KT>, <@U01KZ61GV5Z> Thank you for the reply.


 Do you mean Native Android and IOS SDK already support this feature?

Do you have any road map or something for the flutter?

Thanks again.

yup, the native SDKs support this functionality. Having flutter support for it is on our roadmap. But it likely will take a bit of time to support it fully, due to the complexity with the Flutter stack.

<@U035D2JRL1M> it's very hard to answer this question for your specific use case.. but generally it should have no problem publishing a 720p video to 100 viewers per core.

there’s certainly the option to do something super minimal, but we’re trying to show the most commonly used capabilities. Perhaps in the future, there’s an opportunity to do something more like:

<https://solidity-by-example.org/>

<https://github.com/TheCactusBlue/livekit-why-does-this-break>

I've created this repository that is a very stripped down version of the original repo, that should theoretically work

something like this happens, which is weird

IMO, the default example project is too closely coupled with the main project

which makes it somewhat painful to test in isolation

I don't see any errors in that log, what's the actual problem?

I keep getting error logs that look like this

This is a warning, not an error. It's a problem with CRA. Sourcemap generation is optional, you can ignore it, disable Sourcemap generation, or Google some of the fixes people have for CRA and Sourcemap generation.

just had another test run, fails to connect on a completely fresh CRA project

are you able to reproduce this? this could mean the TURN username/password used was incorrect.

thanks, this is tricky, I finally find out, nginx as reverse-proxy, config `proxy_read_timeout` ,actually it was [to keep alive interval], but I took it as request response read timeout thing, so I configured to 3s, then the SignalClient has to reconnect constantly, caused diff UFrag or something between client and server, here the error log it was result of the problem, not the root cause of the problem

<@U031GJTM8F2> Are you using Turn behind Nginx? If yes how did you setup?

<@U03EFM1KPNH> sorry for late, No I'm not, I can not see why would we have to put turn behind nginx. I put livekit-server behind nginx indeed

now I understand why, back then I didn't know that turn exposes more ports and the client selection was random and thought to proxy it. Of course it didn't work :man-facepalming:

you don't need to keepalive from the client side. the server is sending keepalives via websocket

thanks, It was reverse proxy who disconnect the connection, fixed, thanks

```// connect to room
await room.connect('<ws://localhost:7800>', token, {
  // don't subscribe to other participants automatically
  autoSubscribe: false,
});```

you are right, in order to actually subscribe to tracks you could listen to `RoomEvent.TrackPublished` and then `publiction.setSubscribed(true)` within the callback depending on your needs

related docs: <https://docs.livekit.io//guides/room/receive|https://docs.livekit.io//guides/room/receive>

<@U02EU2ZPTQU> Yeah, I'm migrating from an old version, and I wasn't sure if either of the examples (the "front page" docs one, or the one in the github project) were up to date, but I thought I would point it out to you guys, since it didn't seem to work or be correct.

oh I see what you mean, that the example should not have it set to false.

yeah you are right it's confusing, I'll remove that option. :pray:

Basically, with it false, the example won't work, unless you also listen to track published and do a subscribe yourself.

<@U01KZ61GV5Z> This link (from the page <https://docs.livekit.io/client-sdk-js/>) appears broken:

<https://docs.livekit.io/client-sdk-js/example/sample.ts>

(I get "Access Denied")

hey Alex, all server and client versions are backwards compatible (starting from protocol 3). It's perfectly fine to upgrade one without upgrading the other.

when we start to drop support for older client SDKs, the server will refuse connection from unsupported clients.

similarly, if the server version is no longer supported by clients, the client will also refuse connection.