LiveKit Community

1. Direct RTSP ingestion is a bit tricky because the IP cameras might be behind NAT and inaccessible. You can publish them to LiveKit today with a process that reads RTSP and converts it (like with GStreamer). In the future we could make this easier with the Go SDK.
2. What are you streaming in? If it's RTSP, it's up to you to publish to multiple rooms

hi David, thank you for you answer.
for 2 . you mean, i create different process with one rtsp input, then push to multiple rooms?

yes, this is possible via `updateParticipant` being called with updated permissions through one of the server-sdks.
<https://docs.livekit.io/server/room-management/#updateparticipant|docs>

image.png

Hey <@U05C2NE9UTS>, the pricing page (<https://livekit.io/pricing>) gives a good overview of the differences between the self-hosted open source version and LiveKit Cloud. One of the biggest differences is the maximum number of publishers and subscribers per room. See the “Scale &amp; Reliability” section of the pricing page for more information.

I don’t understand what you mean by that part of your question.
&gt; in the following architecture, what is the equivalent ‘LiveKit Meet’ component in WebRTC?
Can you try rephrasing it?

If you want to host LiveKit yourself, this page is the place to start: <https://docs.livekit.io/oss/deployment/>

Thanks <@U03RM7G8Q6A>,
I'm trying to understand for a 1:1 communication architecture like this (with kiit bot)
where we have a "BOT" that implements webrtc client and joins a session. what is the role of the "LiveKit Meet" service? what does it do? is it an implementation of an equivalent WEBRTC server? or is it something else.

more over, I'm trying to understand the overall architecture.
what's needed to deploy such an architecture on a  cloud platform and be able to scale it.

I understand now :bulb:
LiveKit Meet is our demo app (test here: <https://meet.livekit.io/>, <https://github.com/livekit-examples/meet|code on GitHub>) which is built with LiveKit Cloud and our client SDK. We used it as a base for the Kitt experiment.

You should check out the Kitt repo for some more informations about the project: <https://github.com/livekit-examples/kitt?ref=blog.livekit.io>

<@U05BC1C5EA1> yes, this is possible to do. You can publish a video track and generate and push video frames from the server side. Today, the KITT demo only publishes audio from the backend and then renders the “face” on the client.

i ws going to try and tie it to a live Some sort of agent that not really super realistic face but cartoony in a webRTC scenario

im trying to find a dev to do this but i dont know what to ask or if this is even possible

I assume you’ve seen this?
<https://github.com/livekit-examples/kitt|https://github.com/livekit-examples/kitt>

to connect to our sip so i was trying to get that moving

<https://www.nvidia.com/en-us/omniverse/apps/audio2face/>

kind of like that but its too much i think (so thinking more of a cartoon based AI to face with realtime transcription)

weirdly sitepal had some less relastic and my thought is to STT other side and get the face to move to the response realtime

ahhh ok, yeah we don’t yet support SIP :slightly_smiling_face:

yup, we’ve been thinking about something similar here for a demo — combine STT/TTS + Audio2Face + Unreal’s Metahuman

im getting a POC up for upcoming demo cause its quite cool and innovative hehe

the hardpart alot of those need processsing time to process that and its not a realtime

<https://www.d-id.com/> is what i found is close to that

Yeah, the whole pipeline needs to be optimized around latency.

At least for it to feel like a human-like conversation.

There aren’t yet audio-to-audio models, so what you really want is Text2Face vs Audio2Face so you can generate the TTS and TTVideo in parallel

in a realtime ( i thought sitepal was closet one)

<@U01KZ0SCDKL> wow I forgot we chatted about this long ago haha not sure if there was any cool stuff since last june

Haha, hope you’ve been well, Johnny! Yeah, we’ve been busy: 

<https://github.com/livekit/agents|https://github.com/livekit/agents>
<https://github.com/livekit/sip|https://github.com/livekit/sip>
<https://github.com/livekit/client-sdk-unity|https://github.com/livekit/client-sdk-unity>

<https://blog.livekit.io/open-source-realtime-multimodal-ai/|https://blog.livekit.io/open-source-realtime-multimodal-ai/>

Nice been trying to get sip working too just trial and error

Does it have ability to realtime mouth sync as words are flowing in from other leg of the call

Been trying to find developers to help POC some examples for our cloud instance but hard to come by.

<@U05BC1C5EA1> for clarity, do you mean you want to sync someone’s speech to the avatar’s mouth movements?

Are you rendering the avatar on the server and sending down as a video frame? Or is the avatar rendered on the client?

yea avatar with wss:// streaming to get close to real/live mouth movements i dont knkow if tech can do this yet (or need a powerful GPU haha)

Hello.When I got some problem when using the exapmle app in client-flutter-sdk.
Here is the scenario i met.
First, a participant join the room and chose not to publish any tracks.
Then another participant joined the room and also chose not to publish any tracks.
It seems that the first participant can not be found in widget.room.participants.values.
I'd like to know how can i get remote participants info through client sdk in this situation

Hello <@U05C2NE9UTS>, I'm not sure if you've seen this post but lots of great information in the blog post: <https://blog.livekit.io/meet-kitt/>

But to answer your question KITT is working through the LiveKit Server and if I recall correctly, just acts as any other participant.

yeah I saw that, but I was wondering about costs and what happens during the stream where is the data transferring from and to?

<@U05C2NE9UTS> there’s a nice little diagram in that post detailing how things are wired up.

Pat is correct, KITT behaves like any other participant. A bit more detail on the components:

<https://github.com/livekit-examples/kitt/tree/main/lkgpt-service|KITT>: is a Go program that is deployed as a standalone service and uses the LK GO server SDK’s webhooks and WebRTC integration on the backend to automatically join a meeting room when a human user creates one.

<https://github.com/livekit-examples/kitt/tree/main/meet|Forked LiveKit Meet>: a NextJS app (which uses LK Components) which presents a Zoom-ish UI for meetings. The original version is <https://meet.livekit.io|here>. We forked it because we needed to integrate a “face”/UI for KITT and didn’t have cycles to generate a video-based face on the backend. This app is also deployed to a server and is what’s served up to you @ <https://kitt.livekit.io|https://kitt.livekit.io>. The aforementioned KITT bot/service/program joins rooms in this app when a human connects.

LiveKit Cloud: LiveKit’s <https://github.com/livekit/livekit|media server> is what’s responsible for routing data and media between session participants. In this case, we used our <https://livekit.io/cloud|cloud-hosted> version, because it’s already deployed and running, but you could deploy your own LK media server (“SFU”) somewhere and use that, too. This server facilitates the exchange/transfer of all media between human participants and KITT.

Let me know if that clarifies things or if you have follow-up questions.

Hey Thanks again,
I'm trying to figure out roughly how much a "call" would cost.
in terms of cost of compute (my own cloud resources) + livekitt cloud pricing to handle the webrtc connections and streaming.

One way to make progress is to start with LiveKit Cloud and simply measure your usage after a month. It’s unlikely you cross the free tier in the first month, and you can back into a cost-per-call (both on your side and the LiveKit side) or project how much your costs will be at a certain number of calls each month.

If your app blows up or goes viral in the first month or two, I’d argue that’s a good problem to have. If that happens, hit <@U02BH3ZHJ0Y> up and we can talk more. :slightly_smiling_face:

<@U05C11HU90R> no. When you cluster LK yourself, the SFUs don’t “mesh” or route media between one another.

<@U01KZ0SCDKL> This then means when I create a room at <http://na.myproject.com|na.myproject.com> (visualization in my previous message) and somebody connects to <http://eu.myproject.com|eu.myproject.com> then he won't receive the stream? He has to connect to <http://na.myproject.com|na.myproject.com>, because the the servers don't route data between each other.

<https://livekit-users.slack.com/archives/C01KZ61HZRR/p1686244279794819>

the key way to think about the difference between LiveKit Open Source (OSS) and LiveKit Cloud (Cloud) is, with OSS, a room/session can only be hosted on one server. You can have as many rooms as you want across many servers, but any particular room is hosted by a single server. With Cloud, a single room can be hosted across many servers simultaneously.

I understand, thank you. I'm on cloud right now (and will stay on it for the MVP), but sooner or later I'll have to switch because the pricing is very expensive ($0.18/gb). So that's the biggest down-side for self-hosted, no Mesh SFU then.

You’re welcome. The price-per-gb does decrease as you scale and many Cloud users get pretty far on the free tier.

<@U01KZ0SCDKL> For now I'll stay on cloud. the 50GB free tier will not be enough, but I have no problem with paying for now. Your project is very cool overall. On the client-side I created my own ui, rewote a lot of your react components to VueJS and it's pretty easy to use overall (the abstractions, etc.).

thank you for the kind words! :heart: cc <@U02BH3ZHJ0Y>

&gt; This then means when I create a room at <http://na.myproject.com|na.myproject.com> (visualization in my previous message) and somebody connects to <http://eu.myproject.com|eu.myproject.com> then he won't receive the stream? He has to connect to <http://na.myproject.com|na.myproject.com>, because the the servers don't route data between each other.
It's true that the media doesn't get routed between servers for OSS. However, if you set it up properly, the user should still be able to communicate with each other.

In OSS, each Room is located on a single media server, but if you use a shared Redis instance (or Redis Cluster), we would automatically route all participants of a Room to the same instance behind the scenes. We have a bit of <https://docs.livekit.io/oss/deployment/distributed/|docs here> on that setup.

<@U01KZ61GV5Z> yes, I read the docs. Setting everything up is no problem for me, I was just interested, because the OSS doesn't support Mesh SFU. But <@U01KZ0SCDKL> clarified that for me + thank you for the additional info. I like the product))

this is not supported but we do have <https://github.com/livekit/livekit/issues/1460|a ticket> tracking it. we'll support this soon

Hi, is it happening when you run the final build or you’re seeing the error inside the editor?

This happens when I run the final build at startup time

Is your host machine Windows x64? (where the editor is running)

Hi <@U022KB72NFJ>, did you manage to find the problem?

```function Stage() {
    const cameraTracks = useTracks([Track.Source.Camera]);
    const screenShareTrack = useTracks([Track.Source.ScreenShare])[0];
  
    return (
      &lt;&gt;
        {screenShareTrack &amp;&amp; &lt;ParticipantTile {...screenShareTrack} /&gt;}
        &lt;div className="land-grid-wrapper"&gt;
            &lt;GridLayout tracks={cameraTracks}&gt;
                &lt;TrackContext.Consumer&gt;{(track) =&gt; &lt;ParticipantTile {...track} /&gt;}&lt;/TrackContext.Consumer&gt;
            &lt;/GridLayout&gt;
        &lt;/div&gt;
      &lt;/&gt;
    );
  }```

:point_up: using the stage from the simple nextjs example

you can filter the `cameraTracks` and `screenShareTracks` that are returned by useTracks. i.e. `cameraTracks.filter((t) =&gt; !t.isMuted)`

Hey Charlie,

LiveKit itself does not encode or decode H.264.

The server acts as a video router, simply sending bytes around and does not decode nor encode the video streams themselves.

The clients are utilizing H.264 en/decoding capabilities provided by the browser and/or operating system.

Thanks <@U01KZ61GV5Z> - makes sense. Any general tips other than 'reach out to the patent pools'?

I'm not sure.. are you distributing H.264 encoders/decoders? Have you looked into OpenH.264 by Cisco? They are covering the licensing costs with that particular implementation.. and it's what Firefox and Chrome are both shipping

Not at the moment. here's <https://github.com/livekit/client-sdk-flutter/issues/201|the issue>

is there any example sip integrations to setup this (i was thinking of janus as the middle but i dont know about SDP exchange and such it seems quite tricky)

<@U02MSKUL03V> it’s long been on my list to build a tvOS integration/demo. Wdyt <@U02MAJLGN65>?

hey, yeah, looks like the name parameter is missing in AudioPublishOptions, will fix it

<https://github.com/livekit/client-sdk-flutter/pull/302>

Wow with the quickness! I'll pull this in and give it a try. Thank you so much for your help! :pray:

Works like a charm! Thank you again!!! :face_with_cowboy_hat:

Hey Minh,

on layer pausing - webrtc clients (libwebrtc) would pause additional layers as needed.

all livekit clients SDK support publisher-side congestion control by pausing layers, including Flutter.

you cannot choose which layers not to publish, but if you use `dynacast`, this process becomes automatic. whichever layers is not being consumed will be paused automatically

if simulcast is disabled, it'll publish the highest layer (i.e. what the LocalTrack represends)

Thank you for your assistance.

Can simulcast and dynacast be used simultaneously? If simulcast and dynacast are used together, will there be any conflicts?

In the scenario where a subscriber is in an unstable network environment, but currently has good enough bandwidth to consume the high-resolution layer, while the publisher only has enough bandwidth to publish the high-resolution layer.

- In that case, will the medium-resolution and low-resolution be disabled since there are no subscribers consuming them? (In this situation, is there a conflict between simulcast, which will publish three layers, and dynacast, which will disable the medium-resolution and low-resolution layers?)
- When the subscriber's network becomes weaker and can only consume the medium-resolution or low-resolution layer, will the publisher automatically resume and publish the medium-resolution and low-resolution layers?

yeah, they can be used together.

dynacast always publish the desired layer and lower, because due to congestion/bandwidth issues, we'll need to serve a lower layer to the end user

According to my understanding, when the bandwidth fluctuates, the pause/resume of the layers will be automatically handled, right?

There are quite a lot of handshakes that need to happen in this flow, so there will definitely be some start up delay. There is probably room for improvement here, but our focus for ingress so far has mostly been features and reliability.

That being said, in out testing with OBS, end to end delay is about 2s. I cannot recall what we get with ffmpeg

rtmp is going to inherently have a higher latency and start up time than using a livekit sdk directly to send the media

your ffmpeg cli probably needs to be tuned.. try `-preset veryfast -tune zerolatency`

you can play with these flags/tuning on the ffmpeg side first, then verify with `ffplay` to ensure it matches your expectations first.

going through Ingress will add a bit of latency there, but that should not be the bulk of what you are seeing today.

Does the input have video? you command line is a bit confusing about this. I was assuming it did not

<@U03H6MNQ9G9> the main focus now is on the music only. But long term we want background video as well.

<@U03H6MNQ9G9> do you have a reference how to use livekit sdk directly to stream content? If this option is significantly faster, then we will go for it.

<@U01KZ61GV5Z> thanks for suggestions! I'll give it a try.

If you are publishing static files, the CLI is probably the simplest way to test it: <https://github.com/livekit/livekit-cli#publish-media-files>

there should be minimal latency (&lt;100ms)

here’s an example of publishing an ogg file using the go sdk: <https://github.com/livekit/egress/blob/main/test/integration.go#L103>

Olivier said (pls thread :thread:) :
&gt; ah it was fixed already

oh sorry.. already fixed.. I guess somehow needs to updat the dep on the ingress project and rebuild the docker image.. doing that worked

Top of tree ingress doesn't include the updated module with the fix yet, but I'm about to update it. You can also do this yourself if you're building main yourself anyway

with vp8, the latency is still really high with the ingress module + whip + bypass_transcoding, is that expected ?

Also, any hints on why I'm seeing ~200ms of latency brower-to-browser ?

how much latency are you seeing? what does your encoding pipeline look like?

With whip, in the range of seconds.. I'm using GStreamer v4l2src ! vp8enc ! rtpvp8pay ! whipsink, I measured the latency of that bit to be under 10ms, so I'm assuming it comes from inside livekit

What's the expected latency for the whip code? (Excluding the Network latency, I'm trying the best case scenario with everything localhost first)

We haven't made too much testing with vp8 so far. Mainly, we've used <https://web.whip.eyevinn.technology/>, and got ~600ms end to end latency with bypass transcoding, with a roundtrip to our datacenter

Something seems off in your use case, we'll try to look at it when we have more time to look at performance

We are also about to adopt a new jitter buffer implementation, which may help, particularly at low bitrate. But that would only come into play if there is packet loss, which is unlikely over localhost...

what kind of latency do you get without the ingress module? if I just feed the stream straight to livekit ?

this is including encode/decode.. so end to end latency

what is your use case? I'm curious. and do you have a latency target in mind?

I wonder if I should just add a livekit signaling module to gstreamer's webrtcsink

I want to send video from a device to multiple receives (web and android).. with latency as low as possible (possibly for emote operation of the device)

I looked at little at your Rust sdk, I guess I can just use the right bit s(the gstreamer webrtcsink is in rust)

I think the Rust SDK is probably your best bet.. GStreamer's webrtcsink technically works, but I'm not sure how much work has gone into performance, congestion control/etc.

Rust SDK is built on top of libwebrtc, so you are using what browsers do for their webrtc implementations

I'm one of the gstreamer maintainers, so the gst part is easier for me :wink: I was thinking of just using the signaling part from your sdk, but using gst for all the media handling

hah awesome! that works too.

yeah! that's a good path forward too. We had been thinking about creating some sort of a livekitsink and livekitsrc for GStreamer. to make these types of workflow easier.

lmk if you are interested in collaborating on that. CC: <@U022KB72NFJ>

btw, we use GStreamer extensively in Ingress + Egress and are huge fans. It's an amazing piece of software, thank you for all your work on that.

is there any doc on how the communication between client and sfus work ? would one use the protocol crate directly or is there something more high level ?

here are <https://docs.livekit.io/internals/client-protocol/|some docs> on the signaling protocol. it's a persistent WS connection and all of that should be encapsulated in <https://github.com/livekit/client-sdk-rust/blob/main/livekit/src/signal_client/signal_stream.rs|signal_client>.

currently rtc_engine is the driver that hooks up signal_client with libwebrtc. but it should be possible to just use signal_client as is.

So I ended up writing a GStreamer webrtcbin signaller to connect to LiveKit, but I'm having two issues. I don' tunderstand what the two data channels are for ? And it seems I'm not getting video decoded by the browser, even though the browser claims to be getting the packets .. Is there a link ? It took me  a while to realise it would time out if I didnt create the data channels

Code is here if you're curious <https://gitlab.freedesktop.org/ocrete/gst-plugins-rs/-/tree/livekit-webrtcsink/net/webrtc/src/livekit_signaller|https://gitlab.freedesktop.org/ocrete/gst-plugins-rs/-/tree/livekit-webrtcsink/net/webrtc/src/livekit_signaller>

It's sink-only for now (so publisher only for you)

oh, I found the problem, I had to declare it as a camera

next question.. how do I get two streams to show up in the demo web app ?

You mean <http://meet.livekit.io|meet.livekit.io>?

I'm testing with the sample from the js sdk

For the JS sample, specifically, but you should be able to just open two instances (browser windows), make sure you’re using unique identities for each user and you’ll be publishing two camera streams (one for each participant).

I want to push two tracks from my GStreamer sink, but receive them in the same browser

Aha, I’ll let <@U01KZ61GV5Z> or <@U022KB72NFJ> chime in on this one

Hey <@U05CYS4TVUH>, yes it seems like the example inside client-sdk-js only subscribe to one video track of a participant. This is the same for the Meet App (Camera + Screenshare).
You’ll need to create a custom js app to be able to draw every video of a participant, this can be done very quickly using <https://github.com/livekit/components-js|components-js >(e.g I think you can use <https://github.com/livekit/components-js/blob/main/docs/alpha-docs/components/TrackLoop.md|TrackLoop>)

are the data channels only for the benifit of the app developer? you're not using it in the signaling ?

Do I need to signal the layers in the AddTrackRequest or should I just ignore that ?

and how do I make livekit understand the connection between the tracks and the m-lines? I tried setting cid and name to the value of the mid in the sdp, but it doesnt seem to help? I also requested the tracks in the same order as the sdp

On your first Q: that’s correct, data channels are for the developer, we use a separate websocket for signaling (you need an external signaling mechanism/channel to establish a WebRTC peer connection in the first place)

Any chance to move the signal_client module into a standalone crate? So I don't have to build against libwebrtc ?

Also, is RTX meant to be supported? LiveKit seems to remove it from its answer

<@U05CYS4TVUH> this looks great! yeah I think we can get the signal_client into the API module so it can be used independently.

We do not support RTX today, does GStreamer webrtc handle the non-RTX case as well?

Yeah, rtx isn't a requirement for GStreamer. I was looking at it as a way to improve the stream quality

(more like surviving mild network impairements)

got it, RTX would be an improvement for sure; though we don't believe it to be a main contributor to quality related issues. If you are seeing any odd behaviors, let us know!

just to confirm, does webrtcbin handle NACKs and retransmit in primary stream?

it sends using a separate SSRC, but inside the same RTP session

hmm, that might be an issue then. when RTX isn't negotiated, we expect the retransmissions to occur with the same SSRC. I believe that's what the RFC calls for and what libwebrtc does

oh, in GStreamer we just disable it if its not negotiated

ah.. ok, so that would mean quality will suffer a bit there..

I just moved the signal_client to livekit-api.
The code changed a little bit compared to the one you used on your repo.
`livekit-api = { git="<https://github.com/livekit/client-sdk-rust>", version = "0.1.0", default-features = false, features = ["signal-client"] }`

Thanks, I made a MR to GStreamer with it ! <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1262>

this look awesome <@U05CYS4TVUH>! Excited to give this a try. we have a number of folks that are looking to ingest using GStreamer.. this would be the best path forward there.

This is def interesting.  Could replace how we’re publishing videos with the cli. 

One more request, any chance to make releases of the livekit-api and livekit-protocol crates to <http://crates.io|crates.io> ?

what do you mean by the connection is unsuccessful?

yes，My device is connected to a 4G network and has not been successful. If I switch to a wired network, it can be successful

maybe check if ipv6 is enabled on your server?

I don't know why, it's okay now. It may have something to do with my network. Thank you.

the key pieces of software behind them are the same LiveKit SFU. so all of the encryption are the same. with self-hosting, you are ultimately responsible for securing that install, including getting the right TLS cert for the signal connection. That would be the primary difference

Thank you!
We are looking at replacing Agora with LiveKit and i just wanted to make sure it is as secure. They have something called "Secure channel encryption" where you initiate the client with an encryption key and a salt. I guess LiveKit does something similar but automatically when the signal connection is established?

<https://docs.agora.io/en/voice-calling/develop/media-stream-encryption?platform=web>

I believe some apps are already running the RN SDK in production. cc <@U01NZKTNE1H> 

ah yeah, react-native has been marked for production for a while now. do you mind letting us know where you see the beta tag so we can remove it?

at the bottom of most client sdks, including the react native one <https://github.com/livekit/client-sdk-react-native|sdk> the section that mentions LiveKit Ecosystem

also can we configure the audio to play from a custom resource something like <https://docs.agora.io/en/voice-calling/develop/custom-audio|this> ?. I'm unable to find it in the docs

right now we don’t support custom audio for the native sdks

we don’t have it on our current short term roadmap

Screenshot 2023-06-30 at 11.24.15.png

I found there is no react native introduction in docs, and Docs is mentioning  React Native is (beta).  And this React Native library was worked fine  in expo before but currently not working. There maybe no reason to not support expo (just kind of lost of opportunity lost) .    <https://docs.livekit.io/references/client-sdks/> <https://github.com/search?q=repo%3Alivekit%2Fclient-sdk-react-native%20beta&amp;type=code>

cc <@U03RM7G8Q6A> docs update for react native

Expo instructions can be found here <https://github.com/livekit/client-sdk-react-native/wiki/Expo-Development-Build-Instructions|https://github.com/livekit/client-sdk-react-native/wiki/Expo-Development-Build-Instructions>

I already did such a kind of introcution, it doesnt work current version ^1.1.2, works fine in ^1.0.0. you should try. maybe the result of this commit. <https://github.com/livekit/client-sdk-react-native/commit/26b4ce40e705c4ebb5d8381ff37817fafce7b681> current implementation is not work with @livekit/react-native-webrtc in expo (it means 1.0.0 works fine without @livekit/react-native-webrtc)

<@U01KZ0SCDKL> Hey! is there a public roadmap available yet as well? Thanks!

<https://github.com/livekit/client-sdk-unity>

The developer preview is already released :slightly_smiling_face:

It's not ready for IOS and Android is it?

Not yet, we’re working on it :slightly_smiling_face:

Hi Team, I am looking for paid support for my project where I am using SQL,NodeJs, Front End Android and React Js. I really apprciate if someone can help for same. I want to learn during this development. Thanks

nope, all the published tracks go through the single publisher PeerConnection

where are you trying to add the watermark?

I want to add watermark to video when recording, the solution that I found seem to can’t be use

there’s a videoProcessor parameter that can be added when creating a video track manually

<https://github.com/livekit/client-sdk-android/blob/dc1e9015f4116a49d195f8764980d682ed259760/livekit-android-sdk/src/main/java/io/livekit/android/room/participant/LocalParticipant.kt#L98C12-L98C12|https://github.com/livekit/client-sdk-android/blob/dc1e9015f4116a49d195f8764980d68[…]in/java/io/livekit/android/room/participant/LocalParticipant.kt>