LoginSign Up
← Back to all sparks
J

Jitsi

MEETINGS
Velocity2.5

Open source video conferencing platform from 8x8 with Jitsi Meet and JaaS cloud hosting.

jitsi.org ↗

Jitsi adds distributed tracing across its backend, after rebuilding transcription from scratch.

open-sourcevideo-conferencingobservabilitytranscriptionwebrtcperformance
◆Current state
Jitsi (Jitsi Meet plus its open-source backend) is an engineering-led project whose blog doubles as its changelog. The newest post covers distributed tracing implemented across the main backend components, generating events in each service so a call can be followed end to end. It follows July's from-scratch transcription architecture replacing the decade-old Jigasi approach, and sits alongside community items like the Google Summer of Code cohorts.
◆Where it's heading
The arc is toward operating the media stack at scale, not just building it: selective audio subscriptions, SSRC rewriting and AV1 addressed efficiency, and now tracing addresses diagnosability. For a project whose users largely self-host, being able to follow a call across services is what turns an intermittent media fault from guesswork into something reproducible. Legacy components keep being shed rather than patched — Jigasi for transcription, and now opaque cross-service debugging.
◆Prediction
Expect the tracing work to surface concrete latency or packet-loss findings in later posts, and the new transcription architecture to carry broader real-time features such as live captions; continued media-pipeline optimization for large meetings is the steady background.

◆Recent moves

  1. 13d ago

    Tracing calls through backend components

    Distributed tracing now spans Jitsi's main backend components, emitting events in each service so a request can be followed across the stack. It is operator-facing rather than user-facing, and it is the diagnostic counterpart to the scaling work — self-hosters get a way to locate faults the media metrics only hint at.

    View source ↗
  2. 1mo ago

    A new architecture for transcription (and more)

    Jitsi is replacing its long-standing Jigasi-based transcription with a new architecture, a foundational rework that positions the project for more capable real-time speech features beyond the 2017 prototype.

    View source ↗
  3. 1mo ago

    Google Summer of Code 2026 – Meet This Year’s Projects!!

    Announcement of the five Google Summer of Code 2026 projects; community and contributor news, not a product change.

    View source ↗
  4. 11mo ago

    Introducing Receiver Audio Subscriptions

    Receiver audio subscriptions extend the existing video ReceiverConstraints model to audio, letting clients select which streams they receive, a real efficiency lever for large calls.

    View source ↗
  5. 1y ago

    GSoC 2025, let’s do this!

    Acceptance into Google Summer of Code 2025; contributor-program news with no product impact.

    View source ↗
  6. 1y ago

    AV1 and more … how does Jitsi Meet pick video codecs?

    AV1 becomes the default video codec in Jitsi Meet, improving compression efficiency. A concrete media-stack upgrade, though it predates the current window.

    View source ↗