- Prerequisites
- Quick start
- Exposing your Jitsi Meet installation
- Recording and streaming support
- Transcription support
- Scaling your installation
- Adding custom Prosody plugins
- Documentation
Jitsi-Meet: Secure, simple and scalable video conferences that you can use as a standalone app or embed in your web application.
This chart bootstraps a Jitsi Meet stack on Kubernetes.
See also jitsi-scaler for a Jitsi Meet stack containing multiple shards.
- A Kubernetes cluster and Helm 3.8+ (3.8+ is required for the OCI install).
- A way to expose JVB media to participants (see the Exposing section below).
- cert-manager, only if you enable TURNS (see the TURNS guide under Documentation).
helm repo add jitsi https://jitsi-contrib.github.io/jitsi-helm/
helm install myjitsi jitsi/jitsi-meet --set publicURL=https://meet.mydomain.comAlternatively, install directly from the OCI registry (GitHub Container Registry):
helm install myjitsi oci://ghcr.io/jitsi-contrib/jitsi-meet --set publicURL=https://meet.mydomain.comOCI releases are signed with cosign (keyless). To verify a release before installing:
cosign verify ghcr.io/jitsi-contrib/jitsi-meet:<version> \
--certificate-identity-regexp '^https://github.com/jitsi-contrib/jitsi-helm/' \
--certificate-oidc-issuer https://token.actions.githubusercontent.comJVB must be reachable by all participants (typically on a public IP), and the best approach depends on your cluster. The chart supports several options: LoadBalancer, NodePort, hostPort (including auto-detected node IP and port ranges), hostNetwork, and ingress TCP/UDP forwarding.
See the Exposing guide for all options and examples.
Jibri lets Jitsi Meet users record meetings and live-stream them (e.g. to
YouTube). Enable it under the jibri key, with options for single-use mode,
multiple instances, local recording, live streaming, and persistent storage.
See the Recording and streaming guide for the full configuration.
The chart supports near real-time transcription (subtitles). The current path is
bridge-based: Jicofo tells JVB which transcription service a conference should
use. JVB streams the audio to opus-transcriber-proxy and the proxy relays it
to a speech-to-text provider (OpenAI, Deepgram, Gemini) or to your own
OpenAI-compatible service. Enabling it takes opusTranscriberProxy.enabled plus
the force_async_transcription Prosody module, which the chart does not switch
on for you.
See the custom AI service guide for the full configuration.
The older Transcriber (Jigasi in transcriber mode) with Skynet or another Jigasi-supported backend still works, but it is deprecated upstream. See the Transcriber and Skynet guide.
Jitsi Web, Jibri, and Coturn are stateless and scale freely via replicaCount.
JVB scales via the OCTO relay feature, which spreads the participants of one
meeting across multiple bridges. It can also follow the load on its own, and it
drains before a bridge is removed so that scaling in costs no calls.
See the Scaling guide for replica settings, the OCTO configuration, autoscaling, draining, and how to test it.
Extend Prosody with extra plugins by mounting them via prosody.extraVolumes /
extraVolumeMounts. Modules from the jitsi-contrib/prosody-plugins collection
already ship in the official ghcr.io/jitsi/prosody image.
See the Prosody plugins guide for an example.
Each feature section above links to its detailed guide. See also:
- TURN and TURNS on port 443
- Security hardening
- Custom CA certificates
- Troubleshooting
- Upgrading
- Sample values files
- All guides
For the full configuration reference, see values.yaml.