← back to blog

Hosting a telegram voice chat for hundreds of members without it falling apart

A telegram voice chat looks simple from the audience side. Someone taps a green icon, a wave animation shows who is talking, people listen. Underneath that, there’s a queue of speakers, a set of admin permissions, a media stream that routes differently than your text messages, and an account that has to stay logged in and stable for however long the session runs. If you’ve only ever joined one, none of that matters. If you’re the one starting it for a group of a few hundred people, all of it does.

What starting a voice chat actually does

When an admin (or a member with the right permission) starts a voice chat in a group, Telegram spins up a live session tied to that group, separate from the text chat itself. Anyone in the group can join as a listener without needing to be unmuted or approved. Speaking is the part that’s gated: by default only admins can talk, but admins can open the floor so any member can request to speak, or hand the mic to specific people directly.

This is different from a Telegram call between two people. A private call tries to negotiate a peer-to-peer connection between the two devices first, and only falls back to Telegram’s relay servers if a direct connection can’t be made. A group voice chat doesn’t do that handshake, because there’s no single “other side” to connect to. It runs through Telegram’s own servers the whole time, the same way the rest of the group’s data does. That distinction matters more than it sounds like, and it’s the reason a lot of assumptions people carry over from private calls don’t apply here.

Listeners scale differently than speakers

The audience for a voice chat isn’t capped the way active speaking slots practically are. Telegram treats the listening side more like a broadcast: people join, get the audio stream, and leave without needing anything from the admin. Speaking is the part that has a real bottleneck, but it’s a practical one, not a hard wall. Every person who wants to talk goes through a raise-hand queue, and an admin has to individually approve or deny each request. With a few dozen people that’s manageable. With a few hundred, if you leave the queue open without moderating it, you end up with a backlog of hand-raises nobody is clearing, and the loudest or fastest requesters end up dominating simply because they got approved first.

The fix isn’t a setting, it’s a person. Someone needs to own the queue for the length of the session: approving, muting people who forgot to lower their hand, and pulling the mic back from anyone who’s supposed to have a fixed time slot. For anything past a small group, that’s a full job on its own, separate from whoever is actually talking.

Recording gives you something to point back to

Admins can record a voice chat to an audio or video file, and it lands in Saved Messages when the recording stops. For a few hundred people, this is worth turning on by default rather than deciding in the moment, because you won’t remember to start it once things are already live and people are talking over each other. A recording also becomes the thing you point to afterward if someone disputes what was said, which matters more in larger groups where not everyone was listening live and the transcript-by-word-of-mouth version drifts fast.

Where a proxy fits and where it doesn’t

Telegramvault runs proxy configuration for a living, so it’s worth being direct about this: an MTProto proxy changes how your device reaches Telegram’s servers for the connection as a whole, but it doesn’t change how the voice chat’s media stream is handled once you’re in it. Since group voice chats route through Telegram’s own infrastructure rather than trying to connect directly to other participants, there isn’t the same IP-exposure risk that a private call carries when peer-to-peer is left on. That’s good news if your reason for using a proxy is about where your account connects from generally, not about hiding your address from other people in the call, because a group voice chat was never going to hand your address to two hundred listeners in the first place.

Where a proxy does matter is upstream of the voice chat: if the admin account is running through a proxy for account safety or regional access reasons, that proxy needs to be stable for the entire duration of the session, not just for logging in. A connection that drops and reconnects mid-session can boot the admin out of the voice chat itself, which is a worse failure than losing a few text messages, because rejoining as a listener doesn’t hand your admin controls back automatically. If you’re hosting through a proxy, treat the voice chat window as a period where that connection cannot be touched, restarted, or rotated.

Keeping the host account stable through a long session

A voice chat with a few hundred members generates a lot of background account activity: people joining and leaving, hand-raises, admin actions, messages in the text pane running alongside it. None of that is unusual for Telegram to handle, but it means the admin account running the session should be one with a settled, consistent login history, not a fresh session on a new device that’s about to get flagged for behaving oddly at the exact moment you need it to stay connected. Telegram’s spam and flood protections aren’t looking at the voice chat specifically, they’re looking at the account’s overall pattern, and an account that logs in from a new location right before hosting a large live session is giving itself worse odds than one that’s been sitting stable for weeks.

This is the part that’s easy to overlook because it has nothing to do with the voice chat feature itself. It’s account hygiene applied to a moment where the account can least afford to get interrupted.

The bot that actually keeps things moving

Most groups running voice chats at this size lean on a bot to manage the parts Telegram doesn’t automate: logging who spoke and for how long, posting a recap in the text chat once the session ends, or queuing up the next scheduled voice chat so members get a notification ahead of time. None of that requires anything exotic, but it does require the bot’s host to be up and reachable for the whole session, because a bot that goes offline mid-voice-chat doesn’t come back and catch up, it just misses whatever happened while it was down. If you’re running that kind of bot on infrastructure you don’t control tightly, the voice chat is the moment its uptime actually gets tested.

What to check before you go live

Before starting a voice chat for a group this size, it’s worth running through a short list rather than assuming Telegram will handle the edges for you: confirm who has permission to unmute themselves versus who needs approval, decide who’s actually watching the speaker queue for the duration, turn recording on before you start rather than partway through, and make sure the admin account’s connection, proxy included, isn’t going to need to move or restart while the session is live. None of this is complicated on its own. It just has to be decided ahead of time, because once a few hundred people are in a voice chat, there’s no good moment to pause and fix it.

If you’re setting up hosting or proxy configuration for a Telegram community that runs regular voice chats, that’s the kind of thing we work on daily. Head back to the telegramvault.org home page to see how we handle it.

Get new guides and videos first — join the Telegram channel.

need infra for this today?