anyone could get ai_coustics works? I only get the error from livekit.plugins.ai_coustics of: “Unable to authorize model use: Model use unauthorized” seems there are a feature gate check, but we are on paid account

Hello everyone, I've been self-hosting LiveKit and running into issues at scale, specifically with egress composite recording. I have an API that creates a room, dispatches an agent, and initializes egress composite. If any of these steps fail, the API raises an error. At scale (~40 concurrent users), the API slows down significantly mainly due to waiting on egress. CPU usage on the egress server peaks at ~30% and memory at ~50%, so resources don't seem to be the bottleneck. What I'm trying to understand: what is the actual bottleneck here? Autoscaling or pre-warming a few egress servers would likely help, but I want to diagnose root cause first. I've considered skipping the await on egress initialization to make the API respond faster, but the issue is that audio recording is required for my use case, so I can't just fire-and-forget and move on. Any guidance appreciated!

how many concurrent calls can a self hosted SFU can handle?

Hey , I selfhosted livekit and running my agents in k8s (each pod spawning x agents).. How i can know much each agent is using the CPU . my cluster resouces are occupying fastly. what is avergae vcpu used by an agent session

has any one used dtln noise suppression? I have integrated it and testing in livekit playground but getting no logs to confirm if dtln is working. Is there any other way to test it?

Hello team I am trying to integrate Whatsapp Calling with livekit. All the configs are done but while creating an outbound trunk with transport as TLS, I am not getting the SIP invite. On the other hand, if I am having transport as Auto/UDP, I am getting the SIP Invite but its not moving forward. Any idea on how to fix this

Hi everyone, I wanted to ask if anyone has deploy an agent in aws agentcore and if so, by any chance do you have a guide? I was thinking to deploy in ecs as stated in the documentation but I'm also considering deployment in agentcore. any thoughts?

Hello all, I have automated my onprem deployment from the infrastructure stand point, but I want to automate the twilio process of purchasing a number and creating the SIP trunk. I saw that twilio has an mcp server, but I have been struggling to find the correct key for it. Anyone have luck with this?

Hey folks, I’m planning to self-host LiveKit and LiveKit SIP server on DigitalOcean and wanted to learn from anyone who’s done something similar. I’m expecting around 4-8 concurrent agent sessions per day, so I’m trying to figure out the right setup for smooth scaling and reliability. Any tips on droplet sizing, scaling strategy, or general best practices for this kind of real-time (WebRTC + SIP) workload would really help. Also curious about any pitfalls to avoid. Thanks!

Hello team I am trying to integrate Whatsapp Calling with livekit. All the configs are done but while creating an outbound trunk with transport as TLS, I am not getting the SIP invite. On the other hand, if I am having transport as Auto/UDP, I am getting the SIP Invite but its not moving forward. Any idea on how to fix this

Hi Guys I have hosted livekit on my server(mumbai) Im having more latency over sip calls My STT, LLM , TTS are fast (Groq ~300ms, Cartesia ~300ms and Deepgram ~220ms), everything up to date , twilio for number No audio packet loss or jitter, No CPU or memory issues On web calls(~1.5s to respond) it works fine but sip calls it takes time like ~2.6s How should i figure out the root cause, i have tried everything...

hey folks for self hosted deployement how are you guys doing the scaling of concurrency and what sort of cmpute instance is needed for 100, 1000, 100k concurrency? can yous help me on this and also what is the recommended vcpu and memory size for one concurrent connections?

Hi, We have a platform-specific video quality issue with LiveKit that's impacting our users and we'd like to escalate it. We're using VP8 + Simulcast, which works perfectly on macOS and Linux. However, on Windows, published tracks show inconsistent quality — dropping to Poor before recovering to Good, creating a degraded and unreliable experience. This is causing friction for our users and we need to understand the root cause. Is this a known issue with VP8 Simulcast behavior on Windows? Are there recommended workarounds (e.g., switching to H.264 or AV1 for Windows clients)? Any guidance would be appreciated.

Hello, Our accountant cannot explain an we paid for LiveKit in 2025 and I cannot find the invoice associated. Can someone urgently reach me out please? Regards

Hey folks 👋 My Startup is Running a Voice AI setup on a single EC2 (c7i.4xlarge – 16 vCPU, 32GB) with LiveKit + FastAPI + Twilio + STT→LLM→TTS + Redis. Trying to hit 50+ concurrent calls, but it starts failing around 10–25 calls. Looking for advice on: • Best way to scale this on AWS (ECS / EKS / multi-node?) ( we need to save cost also) • How to scale LiveKit + real-time pipelines properly • Any known bottlenecks in this kind of setup Would really appreciate any pointers

Hi team! 👋 We're migrating from LiveKit Cloud to self-hosted on AWS EKS (Terramate + OpenTofu, Helm, Karpenter) and would love some guidance before committing to a pattern. A few questions:

1. Helm index looks stale.

Chart.yaml

on

livekit-helm

master is at

livekit-server

1.11.0 /

egress

1.12.0 /

ingress

1.5.0, but

<https://helm.livekit.io>

only serves 1.9.0 / 1.8.4 / 1.2.2 (egress last published May 2024). Is a re-publish coming, or should we consume the chart from a git ref?

2. Recommended EKS pattern?

hostNetwork: true

+ dedicated node group, or STUNner /

l7mp/livekit-operator

in front? Any sharp edges with Karpenter consolidation or NLB UDP for TURN?

3. Backend: has anyone validated distributed mode against ElastiCache Valkey (Redis-protocol fork), or should we use real Redis?

4. Egress to S3 via EKS Pod Identity / IRSA — any gotchas vs. static creds?

Thanks!

Hi everyone! I am part of a team putting together an educational mobile app for iOS and Android. As part of that, we are self-hosting LiveKit and using the app-side SDK to operate real-time avatars for learning. It works, but we’ve hit a lot of snags with performance. A lot seems like we just aren’t totally understanding how to get the app to talk to the server correctly. Everything looks like it “should” work, but there are frequent unexplained gaps and sometimes it feels like it works “by accident.” Has anyone been through this route and able to advise? We’ve factored the avatar part of the app into an internal SDK (and deployment), so the LiveKit part is pretty well isolated. As expected, a lot of APIs are involved to get the speech, ASR etc. to work.

Hi everyone, I hope you’re doing well. We are currently looking for an experienced LiveKit / WebRTC developer or infrastructure specialist to support our backend developer with our real-time video setup. Project: 1:1 video calls Target regions: Europe, United States, and Latin America Current server locations under evaluation: Amsterdam and Miami We need support with: • LiveKit setup and architecture • Backend integration • TURN configuration • Latency optimization • Monitoring and reliability • Server/network structure for stable real-time video traffic Ideally, we are looking for someone with hands-on experience in LiveKit, WebRTC, SFU infrastructure, and low-latency RTC systems. Feel free to send me a DM if this sounds like a good fit. Thank you, and looking forward to connecting!

Hi LiveKit team, We're sizing a self-hosted LiveKit deployment on EKS for a voice AI use case — S2S agents handling outbound calls in Spanish, ~3–5 min average duration, expected peak concurrency around 8–10% of daily volume. A few questions where your guidance would help us avoid over- or under-provisioning: 1. 10K calls/day baseline. For a steady-state of ~10K daily calls with the profile above, what's your rule of thumb for SFU pod count, agent worker pods, and node sizing (CPU/RAM, NIC)? Any specific guidance on media server CPU per concurrent participant for AI agent calls vs. human-to-human, given the audio is essentially 1:1 (agent ↔️ caller) but with continuous TTS/STT load? 2. Scaling to 100K calls/day. When customers move from ~10K to ~100K daily, do you typically see linear scaling of the SFU tier, or do bottlenecks shift elsewhere (signaling, TURN, Redis, ingress/egress, agent dispatch)? Any architectural patterns you'd recommend at that scale — multi-region, sharded node pools, dedicated agent pools per use case? 3. Dedicated K8s cluster. At what scale do you see customers move LiveKit onto a dedicated cluster vs. running it alongside other workloads? We currently share an EKS cluster with the rest of our platform (Cilium CNI in ENI mode with kube-proxy replacement, IRSA, Kyverno, Tetragon). Are there isolation or perf reasons — kernel networking, CPU pinning, NIC saturation, noisy-neighbor on UDP/RTP — that push toward a dedicated cluster earlier than raw call volume alone would suggest? Happy to share more detail on the agent stack, codec choices, and traffic profile if useful. Thanks, Ángel

Hi everyone, I hope you're doing well. We are currently looking for an experienced LiveKit / WebRTC developer or infrastructure specialist to support our backend developer with our real-time video setup. Project: 1:1 video calls Target regions: Europe, United States, and Latin America Server locations: Amsterdam and Ashburn, USA Infrastructure: Dedicated bare-metal servers — AMD EPYC 9274F (10Gbps unmetered) We need support with: • LiveKit setup and architecture • Backend integration • TURN configuration • Smart routing logic for 1:1 video calls (latency-based and geo-based routing to connect both participants to the optimal server before the call starts) • Latency optimization • Monitoring and reliability • Server/network structure for stable real-time video traffic Ideally, we are looking for someone with hands-on experience in LiveKit, WebRTC, SFU infrastructure, low-latency RTC systems, and multi-region server routing. Feel free to send me a DM if this sounds like a good fit. Thank you, and looking forward to connecting!

I have setup livekit K8s cluster in Rancher provisioned downstream cluster in namespace livekit livekit setup has 3-livekit pod running as daemons on three nodes, redis-master, two redis replicas , three sentient pods master and replicas , In same cluster my backend application running on separate namespace web apps, architecture vise I am using client ——-> Loadbalancer (sni based routing ) for livekit.example.com(wss) and livekit-nturn.example.com (turn url) to backend livekit node on 8443(nginx) ——> livekit 7880. and backend is connecting directly to the livekit via service url something like backend configMap: livekit_info: http://livekit.livekit.svc.cluster.local:7880 livekit-apikey livekit-secretkey firewall port: opened from external: internet ——> livekit node on 3478 and 50k -60k UDP internet ——-> loadbalancer 443 TCP for was SNI routing configuration: firewall Natting public ip — livekit node private ip all three node configure as node_ip: public ip in the config.yaml as turn for Ice candidate information share to the client for media flow. Internal: rest all internal ports are open and services are able to connect. Issue: when I am trying to make A call from user A to userB it's showing connecting calls are not established. Error:

Signature_valid is ::1

11:51:17 | 200 |   267.43434ms | 127.0.0.6 | POST | /_mms/v2/auth/notification/room | -

lib is already initialized: memory manager = 0x7faf64a80910, threading manager = 0x7faf64a809d0, logging_provider = 0, crypto_systems = 0x7faf64a85b20

Signature_valid is ::1

11:51:18 | 200 |   20.429749ms | 127.0.0.6 | POST | /_mms/v2/auth/room/endRoom | -

{"file":"/app/pkg/models/room_service.go:63","func":"mms/pkg/models.(*RoomService).LoadRoomInfo","level":"info","msg":"Requested room does not exist for the given roomId: !iJNYbtwJTjbjbURXuJ:livekit.example.com","time":"2026-05-11T11:51:19Z"}

{"file":"/app/pkg/models/room_service.go:302","func":"mms/pkg/models.(*RoomService).LoadRoomWithMetadata","level":"info","msg":"load room info error from server for the given roomId: !iJNYbtwJTjbjbURXuJ:livekit.example.com","time":"2026-05-11T11:51:19Z"}

{"file":"/app/pkg/models/room_auth.go:163","func":"mms/pkg/models.(*RoomAuthModel).IsRoomActive","level":"info","msg":"The room information: requested room does not exist","time":"2026-05-11T11:51:19Z"}

{"file":"/app/pkg/models/room_service.go:63","func":"mms/pkg/models.(*RoomService).LoadRoomInfo","level":"info","msg":"Requested room does not exist for the given roomId: !iJNYbtwJTjbjbURXuJ:quartz1.service421.com","time":"2026-05-11T11:51:19Z"}

{"file":"/app/pkg/models/room_service.go:63","func":"mms/pkg/models.(*RoomService).LoadRoomInfo","level":"info","msg":"Requested room does not exist for the given roomId: !iJNYbtwJTjbjbURXuJ:livekit.example.com","time":"2026-05-11T11:51:20Z"}

{"file":"/app/pkg/models/room_service.go:302","func":"mms/pkg/models.(*RoomService).LoadRoomWithMetadata","level":"info","msg":"load room info error from server for the given roomId: !iJNYbtwJTjbjbURXuJ:livekit.example.com","time":"2026-05-11T11:51:20Z"}

{"file":"/app/pkg/models/room_auth.go:163","func":"mms/pkg/models.(*RoomAuthModel).IsRoomActive","level":"info","msg":"The room information: requested room does not exist","time":"2026-05-11T11:51:20Z"}

In the livekit daemon pod seems all 200 response not sure what is getting wrong.

Hi everyone, We recently migrated from LiveKit Cloud to self-hosted LiveKit on Azure. Setup: • D8as_v5 VM (8 vCPU, 32 GB RAM) • LiveKit + SIP + Redis + Egress + CoTURN • Room composite Egress recording to Azure Blob • Docker Compose • 3–5 Egress replicas Calls/audio work perfectly, but during concurrent load some recordings fail with: • "can not accept request" • "Start signal not received" • "no response from servers" After scaling Egress replicas and increasing shm_size, stability improved. Question: • Is ~35–40% CPU usage per active room composite recording normal? • Roughly how many concurrent recordings should this VM handle safely? • Is dedicated Egress infrastructure recommended for production? On LiveKit Cloud, much higher concurrency worked for us at lower cost, so trying to understand best practices for self-hosted scaling. Thanks!

We have raised a PR request to merge our code so that our ASR and TTS models can be listed on LiveKit. How to get a status update on the merge request?

@all Hi we need to run 50 concurrent voice calls for livekit self hosted server and agents ( stt tts llm ) external apis pipeline on aws we tested for 4 cpu 10 calls running what setup or system design we need ..please anyone guide

We have self hosted a livekit deployment on azure need to know is there any good did number and sip provider for indian numbers ?

WEBRTC Developer Needed for Server Review / LiveKit self host Setup Optimization We are looking for a WebRTC developer to help us with a server review. We currently have 2 LiveKit servers, one in Amsterdam and one in Miami, and we need support with fine-tuning, optimization, and setup improvements. We pay hourly. The first paid hour would be a meeting together with our backend developer. The goal is to review the current setup together, improve the server configuration, and help our backend developer better understand everything so he can maintain and manage the setup properly in the future. Please contact us if you match these criteria. Thank you.

Hi everyone, we’re debugging an outbound SIP calling issue and would appreciate help. Our setup: - We are using a self-hosted LiveKit server. - We are also using a self-hosted SIP server. - The SIP trunk is configured to call a Twilio PSTN number. What works: - The same SIP trunk credentials work correctly in the LiveKit cloud SIP server. - In our self-hosted setup, the LiveKit room and SIP participant are created successfully. - Media negotiation also starts, so the RTC side appears healthy. What happens in our self-hosted setup: - When we trigger a call, the mobile number rings. - In some cases, the phone receives multiple incoming call attempts. - Each attempt ends automatically after a short time. - The logs show Twilio returning a 407 auth challenge. - After that, the SIP invite fails and LiveKit eventually returns request timed out / 408. What we observed in logs: - SIP participant creation starts successfully. - Twilio sends 407 during the SIP invite flow. - The invite is then canceled / fails. - LiveKit reports the SIP participant request timed out. What we have already checked: - The trunk username is configured correctly. - The same trunk credentials work with the cloud SIP server. - So the issue seems specific to our self-hosted SIP server or its auth flow. Question: - Is there anything special we need to configure in the self-hosted SIP server to handle Twilio’s 407 challenge correctly? - Could the self-hosted SIP server be failing to send the correct password or digest auth response? - Why would the call be retried multiple times and create multiple ring attempts on the mobile? If anyone has seen this behavior with self-hosted LiveKit SIP + Twilio, any pointers would be appreciated.

which is the livekit stable version for selfhosting it

https://community.livekit.io/t/is-audio-screen-capture-through-screen-sharing-possible/1269%7C is there anyone that could answer my question/give me directions

Hi everyone, I’m working on LaSuiteMeet (https://github.com/suitenumerique/meet) and currently running a large-scale LiveKit deployment for the French administration. Feel free to reach out if you have advanced questions around self-hosting or large deployments. We’re also planning to open source some internal tooling we’ve been building to monitor and operate our infrastructure. In parallel, we’re working on a fork of livekit-sip to allow LiveKit instances to connect with RoomKit hardware with @proud-processor-44533. A first version is almost ready. I’ll be in New York from June 19th to 26th and would be happy to meet members of the LiveKit community there, feel free to ping me. And if you find the project interesting, giving us a star on GitHub would really help us gain visibility and grow the community !