The Jitsi on AWS by FOSSonCloud pattern now ships Jitsi Meet build 11031, up from 10888 — three build increments over roughly three months.
What changed
10888 (March 2026)
Breakout-room message filtering support, RTC stats collection, and full user context passed through the external API.
10978 (May 2026)
MID RTP header extension support for better multi-stream handling, plus a Prosody module for validating MUC (multi-user chat) resources.
11031 (June 2026) — the largest of the three
- A transcription translation toggle for live captions — captions can now be translated, not just transcribed.
- A substantial accessibility pass: better keyboard navigation, ARIA labels across inputs and video-quality controls, and improved error-message association.
- A Salesforce integration overhaul.
- AI-powered virtual backgrounds rebuilt on TensorFlow.js body segmentation running through WebGPU/WebGL, which should mean better performance and quality than the previous implementation.
Why it matters
Translated live captions are the standout feature. Real-time transcription already made meetings accessible to deaf and hard-of-hearing participants; adding translation on top means a distributed team can hold a meeting where participants read along in different languages. For organizations that were previously choosing between “everyone speaks English” and “hire an interpreter,” that’s a meaningful third option.
The accessibility pass is the change with the widest reach and the least visibility. Keyboard navigation and ARIA labeling across inputs and video-quality controls determine whether someone using a screen reader can join a call at all, or whether they get stuck on an unlabeled button. If your organization has accessibility obligations — public sector, education, or anywhere with a VPAT to maintain — this is the release that improves your position, and it’s worth re-running whatever accessibility audit you use.
The virtual background rework matters mostly for client-side performance. Moving body segmentation onto WebGPU/WebGL shifts the work to the GPU, which means less CPU contention on participants’ machines during a call — fewer people whose laptop fans spin up and whose video quality degrades because background blur was eating a core.
The RTC stats collection and full user context in the external API are less visible to end users, but they’re the pieces that matter if you’re running Jitsi as part of a larger integrated stack — they’re what let you attribute call quality data and identity correctly across systems.
What you should do
- Treat this as a routine update. There are no breaking changes or required migration steps.
- Enable and test translated captions if you have multilingual participants. Confirm the languages you actually need are covered, and set expectations — machine translation in a live meeting is useful but not perfect.
- Re-run your accessibility audit if you maintain one. The keyboard navigation and ARIA work should improve your results, and you want that documented.
- Ask participants to re-test virtual backgrounds, especially anyone who previously turned them off because of performance. The WebGPU/WebGL implementation may make them usable on machines where they weren’t.
- Review your Salesforce integration if you use it — the integration was overhauled, so verify your configuration still behaves as expected rather than assuming it carried over.
- Pick up RTC stats if you’re doing call-quality monitoring. Combined with full user context through the external API, you can now correlate quality data with identity properly.
Version details in this post come from Jitsi Meet’s own release notes. If you hit anything upgrading the pattern, open an issue on GitHub.
— FOSSonCloud