Jitsi Meet vs Nextcloud Talk

Jitsi Meet

The engineering is real and published: a selective forwarding unit shown distributing over a thousand streams on a single quad-core server, desktop sharing, and a Last-N mode that protects constrained connections. But we found no public information on breakout rooms, waiting rooms or per-participant moderation, and the vendor states that on the public instance all users are moderators — a room where everyone holds every button is a room nobody controls.

Nextcloud Talk

Breakout rooms and a webinar lobby exist, and scale is honestly published — 6-10 participants for a private setup, thousands per call at enterprise. But the evidence says nothing at all about screen sharing, mute-all or other moderation controls, or adaptive behaviour beyond the candid note that a private setup only reaches 20 people 'if all participants have a good network connection' — silence is information, so I sit at the anchor, not above it.

Jitsi Meet

Encryption is documented as mechanism, not slogan: DTLS-SRTP on all media by default, optional end-to-end encryption whose costs are named plainly (chat and polls stay outside it), key handling explained as a shared passphrase, and a candid statement that the outer layer is stripped at the videobridge while packets live only in memory — all in open source we can audit. What holds it back is admission: it rests on a shared password, and we found no public information on lobby admission or joining gated by our own single sign-on.

Nextcloud Talk

Admission control is real (webinar lobby, and a recording-consent prompt), and the vendor states plainly that 'Nextcloud employees never gain access to your data as we do not offer hosting' — a stronger can/cannot-see statement than most — with open source so the claims can be inspected. But there is no documented end-to-end encryption, no key handling, no SSO-gated joining and no per-meeting passcode mechanism; 'multi-layered encryption' is a marketing phrase, not a documented mechanism, and this criterion is judged on mechanism.

Jitsi Meet

Anyone can join from a browser with no install and no account, guests arrive by link, and mobile apps ship through the App Store, Google Play and F-Droid. But we found no public information on an accessibility statement, live captions, keyboard or screen-reader testing, or dial-in numbers — only an undocumented per-room dial-in information page.

Nextcloud Talk

Web, desktop and mobile clients, phone dial-in via the SIP bridge, and live transcription with translation are all evidenced — genuine reach. But there is no accessibility statement anywhere in the evidence, no keyboard or screen-reader claim, no confirmation guests join by link without an account, and the dial-in carries no numbers because it rides on the customer's own SIP setup; that absence keeps it at the mid-anchor rather than 8.

Jitsi Meet

The minimisation defaults are the best I have seen: rooms are ephemeral, chat and speaker stats are destroyed when the meeting ends, no names or e-mail addresses are retained, cloud recordings leave the servers within 24 hours, and the software ships with no preconfigured analytics engine, so no behavioural scoring sits waiting to be switched on. The gaps are mine to name: we found no public information on participant notification or consent when a recording starts or on any audit trail of recording access, the storage region is unstated and recordings head to Dropbox, and with every user a moderator on the public instance there is no hierarchy of who may record.

Nextcloud Talk

A recording-consent prompt in meetings, transcription AI that is 'completely optional', and recordings that live on infrastructure the works council's employer controls rather than a vendor cloud. But nothing on who may record as a matter of policy, no retention or deletion rule for recordings or transcripts, no access logging, and two features that are personal data about everyone in the room — participant CSV export during calls and AI call summaries — come with no evidenced off-switch for co-determination; attendance export without a documented kill switch is exactly what I object to.

Jitsi Meet

Deployment genuinely is our choice: the whole stack is open source and self-installable in as little as fifteen minutes, with documented API access, an embeddable external API, tenant naming and XMPP/REST control interfaces. What is missing from the captured pages: we found no public information on calendar integration, single sign-on, automated provisioning or webhooks for meeting events.

Nextcloud Talk

Deployment is genuinely ours: fully on-premises, open source, even airgap-capable, which for this criterion is the strongest form of 'run it yourself' the anchors describe. Around it sit webhooks for bots, a Pexip integration, SIP bridge and an Outlook add-in — but SSO, SCIM provisioning and an SDK are simply not evidenced, and I cannot credit capabilities the evidence does not show.

Jitsi Meet

This buyer weighs sovereignty heavily, and the hosted offering fails it: the contracting entity is 8x8, Inc. in California, disputes are litigated in New York, meet.jit.si analytics run through the US services Amplitude, Datadog and Crashlytics, and we found no public information on where media and metadata are processed. The open-source self-hosted path is the redemption — on our own European servers the vendor's regions disappear from the chain — but that is a path we build ourselves, not a commitment the vendor publishes.

Nextcloud Talk

The pipeline marked residency unknown because the vendor does not state one — but the reason is that Nextcloud hosts nothing: media and metadata are processed on named customer infrastructure, in an airgap if we wish, and the vendor states no user data is transferred to it. The contracting entity is demonstrably European — Nextcloud GmbH, Stuttgart, HRB 227086, German VAT ID — though I dock it for no published certification or subprocessor documentation, and the vendor-operated public STUN server touching call setup without a stated safeguard.

Jitsi Meet

The published price for the product itself is "completely free", open source, with no tiers to decode, so a self-hosted deployment carries no vendor invoice to miscompute. But we found no public prices for the Jitsi as a Service hosted offering, and the terms let 8x8 require users it judges excessive to "transition to a paid service", for which we found no public price.

Nextcloud Talk

Three tiers are publicly priced per user per year, net of VAT, with volume steps from 100 to 200 users and published participant limits — respectable. But the High Performance Backend, without which there are no webinars and no phone dial-in, is offered only as a 30-day trial with no public price, and no minimum term appears anywhere; a buyer cannot compute the invoice for a setup that actually needs those features.