whats-best.ai
Search Sign in

Video Conferencing · head-to-head

Jitsi Meet vs Nextcloud Talk

Jitsi Meet

Rest of world

Panel rating

Sovereignty: 2 of 4 dimensions proven

Full evaluation →

Nextcloud Talk

EU-Made

Panel rating

Sovereignty: 1 of 4 dimensions proven

Full evaluation →

The written short answer is being updated after a re-evaluation. The scores below are current.

Read this comparison as one judge. Each weighs the same scores by what they care about.

The IT Administrator

Gets the call when the board meeting will not start. Wants capacity published rather than promised, room systems that work, SSO, and diagnostics after a bad call. Judges a product by its worst meeting, not its demo.

Jitsi Meet

This judge's pick

Nextcloud Talk

Criterion by criterion

The meeting itself

Jitsi Meet

The published load tests are exactly what I want before a board call: over a thousand streams at roughly 550 Mbps for about 20% CPU on a quad-core Xeon, on a Selective Forwarding Unit that relays rather than mixes, with Last N mode for large rooms. What holds it back is the bad day — on the public service all users are moderators with soft controls, we found no public information on breakout rooms, waiting rooms, room-system support or behaviour on mobile networks, and the hosted demo carries no uptime commitment and is expressly not meant for production use.

Nextcloud Talk

The capacity honesty is rare and I credit it — '6-10 participants, up to 20 if all participants have a good network connection and do not use video' for private setups, 'thousands' with the enterprise backend — but screen sharing, mute-all, adaptive behaviour on a bad connection and post-call diagnostics are nowhere on this sheet, and the webinar lobby that would serve as a waiting room is gated behind the HPB.

Encryption & meeting access

Jitsi Meet

This is documentation I can defend to security: DTLS-SRTP on all traffic by default, optional end-to-end encryption via a shared key with the trade-offs named plainly (chat and polls sit outside it), and the pages state outright that the outer encryption layer is removed at the videobridge while packets live in memory only and are never persisted — all under Apache 2.0, so the crypto is inspectable. We found no public information on SSO-gated joining, a lobby, or participant identity verification, which keeps it off the top.

Nextcloud Talk

All I get on crypto is a marketing-grade 'multi-layered encryption' line on a feature list — no statement of what meeting media is protected against, no key handling, no E2EE for calls, no SSO-gated joining. Self-hosting means the vendor never touches my data, which is real, but the cryptography itself is undocumented on this sheet, and admission control is a lobby that requires the HPB.

Reach & accessibility

Jitsi Meet

Browser-first over WebRTC with end-to-end encryption and Picture-in-Picture working in the browser, guests join by link with no account, mobile apps ship on all three stores including F-Droid, a per-meeting dial-in information page exists, and speech-to-text transcription is published. But we found no public information on an accessibility statement, screen-reader or keyboard testing, listed dial-in numbers, or low-bandwidth behaviour, so I cannot put a worst-case guest in front of this with confidence.

Nextcloud Talk

Web, desktop and mobile clients are confirmed, phone dial-in exists via the SIP bridge (needs HPB), and live transcription with translation is a genuine reach win. But no dial-in numbers are listed, guest-by-link is only implied by 'public video conferencing', and there is no accessibility statement, keyboard-operation or screen-reader evidence at all — silence I score as absence.

Recordings, retention & admin control

Jitsi Meet

The destruction posture is unusually strong: rooms are ephemeral, chat and speaker stats are destroyed when the meeting ends, names and emails are not retained, and cloud recordings leave for the destination you indicate (currently Dropbox) with anything still on the servers deleted within 24 hours. What an organisation needs around that is absent from the captured pages — we found no public information on admin policy over who may record, participant notification or consent prompts, storage region, an access audit trail, or a retention rule for transcription.

Nextcloud Talk

Recording consent in meetings and participant CSV export are real co-determination-adjacent features, and because recordings live on infrastructure I run myself, the storage region is whatever I choose. But the evidence gives me no admin policy over who may record, no automated retention or deletion for recordings or transcripts, and no access audit trail — a works council could not read its rights off these pages.

Integrations & deployment

Jitsi Meet

Deployment is genuinely the customer's choice — fully open source under Apache 2.0, a Docker or quick-install path in as little as 15 minutes, control over XMPP and a REST interface, an embeddable external API, and a hosted JaaS option with tenant domains. That side is as good as it gets; but for the daily stack we found no public information on calendar integrations, SSO, or room-system support, and the public server is explicitly for testing rather than production integrations.

Nextcloud Talk

The self-hosted story is the best in class — fully on-premise, 100% open source, airgap-capable, with hosted-HPB via partners as fallback — and the Pexip bridge is real room-system interop, with webhook-based bots. But the evidence is silent on SSO, SCIM and any SDK, and the only calendar story is Nextcloud's own apps; for a platform I have to bolt onto an existing identity stack, that silence matters.

European sovereignty

Jitsi Meet

For the vendor-run services the captured pages give me nothing to stand on: the contracting entity is 8x8 in the United States, disputes go to New York courts, data residency is unstated, and the named analytics services (Amplitude, Datadog, Crashlytics) come with no published regional safeguards. The one real answer is structural — the software is fully open source and self-hostable, so media and metadata can run entirely on our own EU infrastructure with no vendor in the chain — but we found no public information on EU hosting, certification, or subprocessor commitments for any 8x8-run offering.

Nextcloud Talk

As a buyer who weighs this heavily: the vendor 'do[es] not offer hosting' and its employees 'never gain access to your data', media stays entirely on infrastructure I control, and Nextcloud GmbH is a German entity with a register entry and Stuttgart address — despite the pipeline's unknowns. It falls short of the top anchors because the vendor-hosted STUN server and the partner-hosted HPB are unnamed and unlocated, and no certification or subprocessor list for the service is published.

Pricing transparency

Jitsi Meet

For the self-hosted software the headline is unambiguous — completely free, no account needed — and the performance page even names the class of server it ran on, so my self-hosting invoice computes to zero. But the same pages advertise a managed JaaS offering for which we found no public information on prices, dial-in and recording costs are unmentioned, and the terms reserve 8x8's right to move excessive users to a paid service at its sole discretion.

Nextcloud Talk

Per-user annual prices for three tiers at two volume points, net of VAT in EUR, from 100 users up — that part I can compute. What I cannot compute is the meeting: the HPB, without which webinars, dial-in and thousands-participant calls do not work, appears only as a free 30-day trial with no price, no SIP rates and no minimum term on this sheet.

Sovereignty, side by side

Dimension Jitsi Meet Nextcloud Talk
Legal entity Incorporated in US Incorporated in DE
Ownership Foreign-controlled Not determined
Data residency Not determined Not determined
Subprocessors Not determined Not determined

Facts, side by side

Only facts both products carry under the same definition — anything else would not be a fair row.

Product · Mobile apps yes · App Store · Google Play · F-Droid1

captured 1 Oct 2026 · Report an error

yes2

captured 15 Sep 2026 · Report an error

Product · Open source yes3

captured 1 Oct 2026 · Report an error

yes · yes2

captured 15 Sep 2026 · Report an error