Nextcloud video conferencing: which option to choose in 2026?
Nextcloud handles your files, your calendar and your contacts without depending on an American cloud. Talk adds messaging and calls. But as soon as a meeting grows past a handful of people, the same question comes back: will it hold? The answer lies in a component many administrators discover too late: the engine that distributes the video.
Is Nextcloud Talk enough for video conferencing?
Yes for small calls, no beyond three to five participants.
By default Talk works peer-to-peer: no central server relays the audio and video. Each participant sends their stream directly to every other one. It is elegant, and it is exactly where it breaks.
In a five-person call, everyone sends four streams and receives four. At ten, it is nine and nine. The load grows as the square of the number of participants, and it is sending that saturates first: consumer connections have far less upstream than downstream throughput.
In practice, Talk’s peer-to-peer tops out at around three to five participants per meeting. With generous bandwidth and processors you can reach about twenty people, but with no guarantee of quality.
For two or three people, native Talk is enough and requires no extra infrastructure. The problem only appears in a team meeting.
What is the High Performance Backend?
It is the official SFU for Nextcloud Talk. It replaces peer-to-peer with a central server that receives one stream per participant and redistributes it to the others.
The change of architecture is the heart of the matter. With an SFU (Selective Forwarding Unit), each participant sends their stream once to the server, which takes care of passing it on. In a ten-person call, everyone sends one stream instead of nine: roughly 90 % less upstream bandwidth.
The High Performance Backend is not a single component but an assembly.
| Component | Role |
|---|---|
| Spreed Signaling Server | orchestrates the connections between participants |
| Janus | the WebRTC gateway that routes media (VideoRoom plugin) |
| NATS | distributes messages between instances |
| CoTURN | relay for participants behind a closed network |
One misconception deserves correcting: the High Performance Backend is not reserved for Enterprise customers. The nextcloud-spreed-signaling signalling server is published under AGPL-3.0 and can be installed freely. Nextcloud sells support and guidance for it, not the right to use it.
On capacity, Nextcloud estimates that a server with 4 cores and 8 GB of RAM carries at least 150 simultaneous users. That is comfortable for a whole small business.
The price is operational: four or five services to install, monitor and keep updated, on top of your Nextcloud.
Should you plug in Jitsi or BigBlueButton instead?
Those are the two classic escapes when Talk is no longer enough. The Nextcloud App Store offers an app for each: Jitsi Integration and BigBlueButton Integration.
The principle is the same in both cases: Nextcloud no longer does the video, it delegates it. The app creates the room and publishes the link; the media lives entirely on the Jitsi or BigBlueButton server.
That solves the scaling problem, but moves it onto three others.
- You run two platforms. Nextcloud on one side, Jitsi or BigBlueButton on the other, each with its own update cycle and monitoring.
- Identity splits in two as soon as the integration fails to carry it into the room.
- Sovereignty depends on hosting. Pointing the Jitsi app at a public instance sends your meetings to third-party infrastructure, which cancels the point of a self-hosted Nextcloud.
BigBlueButton keeps a real advantage: it is designed for teaching, with a whiteboard, polls and breakout rooms. If your need is pedagogical, it remains relevant. We have set out what each of these stacks really costs, as well as the substantive differences with Jitsi and BigBlueButton.
The five options compared
| Native Talk | Talk plus HPB | Jitsi | BigBlueButton | Vuisio | |
|---|---|---|---|---|---|
| Media architecture | peer-to-peer | SFU (Janus) | SFU | SFU | SFU |
| Beyond 5 participants | no | yes | yes | yes | yes |
| Extra services to run | none | 4 to 5 | a separate platform | a separate platform | 1 |
| Nextcloud identity | native | native | depends on the app | depends on the app | native |
| Licence | AGPL-3.0 | AGPL-3.0 | Apache-2.0 | LGPL-3.0 | AGPL-3.0 |
What Vuisio changes in this equation
Vuisio brings together what the other options force you to trade off: the scaling of an SFU, the native Nextcloud identity, and one service to run instead of five.
The Vuisio app for Nextcloud (integration_vuisio, AGPL-3.0) follows the same logic as Jitsi and BigBlueButton, that of an external media server. But it closes the three compromises above.
Identity does not split. Participants join the meeting under their Nextcloud name, and whoever created the room automatically becomes its moderator. That role is sealed by a cryptographic signature on the server side, so it cannot be usurped by a participant tinkering with their link.
One service to maintain. Where the High Performance Backend means running Janus, NATS, the signalling server and CoTURN over time, Vuisio comes down to the SFU. That is the heart of the trade-off, and it is operational more than technical.
No media passes through Nextcloud. The app embeds the Vuisio client in an iframe; audio and video travel directly between participants and your SFU. Nextcloud handles only identity and starting the meeting.
Installation takes three clicks from the App Store, followed by three settings on the server side: the rooms API token, the role signing secret, and the origin allowed for the iframe. All of it is published under AGPL-3.0, auditable, and works as well on your own infrastructure as in a cloud hosted in France.
How to choose by size
The question is not “which is best” but “when do I need more”.
- Fewer than 5 participants, occasional use: keep native Talk. Adding an SFU would be complexity for nothing.
- A team of 10 to 150 people, already comfortable with operations: the High Performance Backend is the most coherent path. You stay in the Nextcloud ecosystem, everything is AGPL, and the capacity is there.
- A teaching need (whiteboard, polls, breakout rooms): BigBlueButton is the best equipped.
- You want the scaling without running five services: Vuisio keeps the Nextcloud identity and brings the surface to maintain down to a single server.
The real trade-off is not technical, it is operational. The High Performance Backend is an excellent solution, provided you have someone to maintain Janus, NATS, the signalling server and CoTURN over time. Many organisations underestimate that cost, install it once, and find themselves two years later with a stack nobody knows how to update. That cost of ownership is precisely what Vuisio sets out to remove.
Whichever option you take, one point does not change: as long as the media stays on your infrastructure or with a European host, you keep control of your meetings. That is the framework we set out in our guide to sovereign video and the GDPR.
Vuisio is open source under AGPL-3.0, free with no time limit, and installs in Nextcloud in a few minutes.
Frequently asked questions
How many participants does Nextcloud Talk support?
In peer-to-peer mode, the practical limit sits between three and five participants per meeting. With generous upstream bandwidth and processors you can reach about twenty people, but with no guarantee of quality. Beyond that, an SFU becomes essential.
Is the High Performance Backend reserved for Enterprise customers?
No, and that misconception is a stubborn one. The nextcloud-spreed-signaling server is published under the AGPL-3.0 licence and can be installed freely by anyone. Nextcloud sells support, guidance and a prepared stack, not the right to use it. You can therefore self-host it entirely without any Enterprise subscription.
What is the difference between peer-to-peer and an SFU?
In peer-to-peer, each participant sends their stream to every other one: the load grows as the square. With an SFU, everyone sends their stream once to the server, which redistributes it. In a ten-person call, that removes roughly 90 % of the upstream bandwidth.
Can Jitsi or BigBlueButton be plugged into Nextcloud?
Yes, the Nextcloud App Store offers an app for each: jitsi and bbb. Nextcloud then creates the room and publishes the link, but no longer handles the media. In exchange you run two separate platforms, each with its own updates.
Which option keeps the Nextcloud identity?
Native Talk and the High Performance Backend use it natively, since they are part of Nextcloud. The Vuisio app carries it into the room as well: participants join under their Nextcloud name, and the room's creator becomes its moderator. For Jitsi and BigBlueButton, it depends entirely on the integration app you choose.
Why choose Vuisio over the High Performance Backend?
For the operational load. The High Performance Backend means maintaining Janus, NATS, the signalling server and CoTURN over time, where Vuisio comes down to a single service. The app carries the Nextcloud identity, seals the moderator role with a signature, and stays under AGPL-3.0, auditable and self-hostable.
Is a TURN server always needed?
Almost always, yes. CoTURN acts as a relay when a participant sits behind a restrictive corporate network, a strict firewall or a symmetric NAT that prevents a direct connection. It is needed for peer-to-peer as for the High Performance Backend, and remains recommended with any external SFU.
Complete guide : BigBlueButton alternatives in 2026: the comparison
