Nextcloud Talk Video Call Server Requirements

Updated 9 November 2026 · Standard formula, shown below

Nextcloud Talk video-call requirements depend on how many people join, whether calls use a high-performance backend, and whether video is relayed through a TURN server. A small, occasional call can work on a modest Nextcloud host, but group calls need more CPU, network capacity, and often separate real-time communication services; there is no single reliable RAM or bandwidth figure for every installation.

What runs a Talk video call

Nextcloud Talk is the communication app within a Nextcloud installation. The Nextcloud server handles accounts, permissions, signaling, and other application functions. The media path—the transmission of audio and video—can involve users connecting directly, a TURN server relaying traffic when direct connections fail, or a High Performance Backend (HPB) that coordinates larger calls.

This distinction matters when estimating resources. A server that is adequate for file sharing and messaging is not automatically sized for large video meetings. Conversely, installing Talk does not mean that every call’s video must pass through the main Nextcloud web server. The actual path depends on the call setup, network restrictions, and the Talk services configured.

CPU, memory, and storage

CPU demand varies with the number of simultaneous calls, participants, video streams, and the media services being used. Calls involving several people generally take more work than a two-person call, especially when the system has to distribute multiple streams. Encryption, connection handling, and other workloads on the same machine also affect headroom.

Memory is needed by Nextcloud, its database and web stack, and any additional Talk services. There is no defensible universal RAM-per-participant rule: deployments differ in software versions, configuration, concurrency, and whether the media backend shares a host. A server with enough memory for Nextcloud’s normal workload may still be constrained by CPU or network capacity during calls.

Video calling does not usually make storage the main sizing concern. Recordings, if enabled through an appropriate integration, can consume substantial storage, but live-call capacity is more directly affected by processing and bandwidth. Monitor the whole host rather than judging readiness by free disk space alone.

Bandwidth and TURN capacity

Network capacity is often the limiting resource. Each participant sends and receives media, and the amount varies with resolution, frame rate, activity, device settings, and network conditions. The server’s outbound traffic can become significant when it relays streams for many participants. Do not treat a client’s advertised connection speed as a guaranteed server-side requirement.

TURN provides a relay path when users cannot establish a suitable direct connection—for example, because of restrictive firewalls or network address translation. A TURN relay consumes network traffic and requires correct public reachability and firewall configuration. It does not, by itself, provide all the coordination features of a Talk High Performance Backend.

Estimate concurrent calls, not just registered users. A 200-user installation with five people on calls at a time differs greatly from one with dozens of simultaneous meetings.
Consider the participants’ networks. Corporate firewalls and mobile connections can increase reliance on relays or reduce usable call quality.
Check both upload and download capacity at the server or relay, along with network limits imposed by the hosting provider.
Leave headroom for normal Nextcloud activity and traffic bursts rather than sizing to an average alone.

When to add a High Performance Backend

Talk’s High Performance Backend is intended to improve scalability and call handling, particularly for larger group meetings. It is an additional service to deploy and maintain, not simply a checkbox that makes an undersized server unlimited. The backend has its own resource and connectivity needs, and the Nextcloud application server still performs its regular duties.

For a small organization with occasional one-to-one or small-group calls, begin with a properly maintained Nextcloud host and measure real use. If calls grow larger or performance becomes inconsistent, review the current Talk documentation for supported backend components and deployment guidance. A separate host or container can make it easier to allocate and monitor resources, but the appropriate layout depends on the organization’s infrastructure.

How to size and test a deployment

Because call patterns vary, sizing should be treated as a capacity-planning exercise, not a guess based on a fixed participant count. Before a production rollout:

Record the expected peak number of simultaneous calls, typical and maximum participants per call, and whether video is commonly enabled.
Identify whether users can connect directly, whether a TURN relay is required, and whether an HPB is appropriate for the expected meeting sizes.
Check that the server and any relay or backend are reachable on the required network paths. Follow the documentation for the exact Talk version and chosen deployment method.
Run representative calls with users on the networks and devices they actually use. Include a busy-period test rather than testing only one short call.
Watch CPU, memory, network throughput, dropped or unstable connections, and application or service logs. Increase capacity or adjust the architecture when measurements show a bottleneck.

Keep Nextcloud, Talk, and their supporting services on compatible, maintained versions. Resource planning cannot compensate for a misconfigured firewall, an unreachable relay, or an unsupported component combination.

Common misconceptions and failure points

A common misconception is that the Nextcloud web server always carries every video stream. Direct media paths may avoid that, while TURN relaying or a particular group-call architecture can put more media traffic on infrastructure you operate. Another misconception is that adding RAM alone will fix poor calls. If CPU is saturated, the network is congested, or users cannot reach the required services, more memory will not solve the underlying problem.

Before buying more server capacity, distinguish between server overload and connection problems on participants’ devices or networks. If only users behind a particular firewall have trouble, review relay reachability and network rules. If quality worsens only during busy periods, compare resource and bandwidth measurements with the call schedule. If the web interface is slow as well as calls, the shared host may be overloaded by unrelated Nextcloud work.

Frequently Asked Questions

How much RAM does Nextcloud Talk need?
There is no universal amount that guarantees good calls. Required memory depends on the Nextcloud workload, concurrent calls, and whether TURN or a high-performance backend runs on the same host. Measure under representative load and preserve headroom.

Can I use Nextcloud Talk without a High Performance Backend?
Small calls may work without one, depending on the deployment and call setup. The HPB is intended for improved handling and scalability, particularly for larger group calls. Check the current Talk documentation to determine which configuration fits your use.

Does every video call go through my Nextcloud server?
No. Media can use direct connections, a TURN relay, or a backend-supported call path. Which route is used depends on connectivity and configuration, so do not assume the main web server carries all media—or none of it.

Why do calls fail behind a company firewall?
Firewalls may block or restrict the network paths needed for direct media or relay services. Review the Talk and TURN network requirements for the installed version, then have the network administrator verify the relevant routes and ports.

How many participants can one server support?
There is no fixed number that applies to every server. Capacity depends on call size, simultaneous meetings, media paths, hardware, and network conditions. Test the expected peak workload and use measurements to set a realistic limit.

Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.