How many meetings can Jibri record at once?

One per Jibri instance. To record three meetings at the same time you run three Jibri containers, each with its own ALSA loopback device, on a machine with enough CPU for three Chrome and ffmpeg processes.

Where does Jibri save recordings?

Inside the Jibri container under /config/recordings, which the Docker setup maps to the jibri folder under your CONFIG directory. Each recording gets its own subfolder with an MP4 file and a metadata JSON file.

Can Jibri upload recordings to S3 or Dropbox automatically?

Jibri runs a finalize script after each recording. Point JIBRI_FINALIZE_RECORDING_SCRIPT_PATH at a script that copies the folder to S3, or have a timer upload finished folders. On AWS, give the instance an IAM role instead of storing keys on it.

Does Jibri work without a domain and SSL?

Only if you tell it to ignore certificate errors. Jibri joins meetings in headless Chrome, which rejects the self-signed certificate an IP-only server uses, and the recording fails with a timeout. Set IGNORE_CERTIFICATE_ERRORS=1 for IP-only servers, or better, use a real certificate.

Jitsi Recording with Jibri (Docker): Setup and Fixes, 2026

Jibri (Jitsi Broadcasting Infrastructure) records a meeting by joining it as a hidden participant in a headless Chrome browser, capturing the screen and the audio, and encoding them to MP4 with ffmpeg. That design explains every requirement below: it needs a virtual sound card, real CPU, and a certificate Chrome trusts.

Requirements

A working Jitsi Docker install. See install Jitsi with Docker.
8 GB of RAM if Jibri shares the Jitsi server, or a separate recorder server.
About 2 CPU cores per concurrent recording.
The snd-aloop kernel module on the machine that runs Jibri.
A domain with a valid certificate, strongly recommended.

1. Load the ALSA loopback module

Jibri captures audio from an ALSA loopback device. Check the module is available for the running kernel:

sudo modprobe -n snd-aloop && echo available || echo missing

On AWS, Ubuntu’s linux-aws kernel does not ship it. Install linux-generic, make it the default GRUB entry, update-grub, and reboot once. The full steps are in Jitsi on AWS.

Then configure enough loopback devices and substreams, load the module, and persist it:

echo "options snd-aloop enable=1,1,1 index=0,1,2 pcm_substreams=8" | sudo tee /etc/modprobe.d/jibri-aloop.conf
sudo modprobe snd-aloop
echo snd-aloop | sudo tee /etc/modules-load.d/snd-aloop.conf

2. Enable recording in .env

ENABLE_RECORDING=1
# Pin it. The Jitsi server and every recorder must agree on this domain.
XMPP_RECORDER_DOMAIN=hidden.meet.jitsi
JIBRI_FINALIZE_RECORDING_SCRIPT_PATH=/config/finalize.sh

gen-passwords.sh already set JIBRI_RECORDER_PASSWORD and JIBRI_XMPP_PASSWORD. If you added Jibri to an older .env, run it again or set them by hand.

Only on a server without a real certificate:

IGNORE_CERTIFICATE_ERRORS=1

3. Start with the Jibri overlay

Download jibri.yml from the same release as your compose file and start both:

cd /opt/jitsi-meet
curl -fsSLO https://raw.githubusercontent.com/jitsi/docker-jitsi-meet/stable-11031/jibri.yml
docker compose -f docker-compose.yml -f jibri.yml up -d
docker compose -f docker-compose.yml -f jibri.yml logs -f jibri

In a meeting, a moderator now sees Start recording in the menu.

4. Record several meetings at once

One Jibri records one meeting. To run more, scale the service with a small override file:

# jibri-scale.yml
services:
  jibri:
    deploy:
      replicas: 3

docker compose -f docker-compose.yml -f jibri.yml -f jibri-scale.yml up -d

Compose v2 honours deploy.replicas outside swarm mode. Give the machine about 2 cores and 2 GB of RAM per replica on top of the Jitsi stack, or move recorders to their own server.

5. A dedicated recording server

For regular recording, run Jibri on a separate machine (on AWS, a c5.large for one concurrent recording or a c5.xlarge for two or three). The recorder connects to Prosody on the Jitsi server on port 5222:

Publish 5222 from the Prosody container on the Jitsi server.
Allow 5222/tcp in the Jitsi server’s firewall only from the recorder. On AWS, reference the recorder’s security group.
Use identical XMPP_RECORDER_DOMAIN, JIBRI_RECORDER_PASSWORD and JIBRI_XMPP_PASSWORD values on both machines.

A mismatched recorder domain is invisible until someone presses Record and nothing happens.

6. Upload recordings to S3

Jibri runs the finalize script inside its container, with the finished session folder as the first argument. Keep it tiny and just mark the folder complete:

#!/bin/bash
# /opt/jitsi-meet/config/jibri/finalize.sh (seen as /config/finalize.sh in the container)
[ -n "$1" ] && touch "$1/.ready"
exit 0

A systemd timer on the host then ships only completed sessions and frees the disk:

#!/bin/bash
REC_DIR=/opt/jitsi-meet/config/jibri/recordings
BUCKET=my-recordings-bucket
shopt -s nullglob
for dir in "$REC_DIR"/*/; do
  [ -e "${dir}.ready" ] || continue
  name=$(basename "$dir")
  aws s3 cp "$dir" "s3://$BUCKET/$name/" --recursive --exclude ".ready" --only-show-errors && rm -rf "$dir"
done

Splitting “mark” from “upload” matters: uploading from inside the finalize script blocks Jibri, and a failed upload would otherwise lose the file. Give the instance an IAM role with s3:PutObject on the bucket instead of storing keys on disk.

Troubleshooting

The three failures we see most, each with a fix, are in Jitsi recording not working:

“Timed out waiting for call page to load”: certificate or network problem.
Recording starts then stops, or has no audio: snd-aloop missing or too few substreams.
The Record button does nothing: recorder domain or password mismatch, or no free Jibri.

Prefer it done for you?

On our platform, recording is a one-time $39 feature per server. It installs the kernel fix, sizes the recorder, and uploads every recording to an S3 bucket in your own AWS account. We also set up Jibri on existing servers as a service.

How to record Jitsi meetings with Jibri