OpenVidu 3.5.0

OpenVidu Platform

Changelog

• ARM support: OpenVidu can now be deployed in ARM-based systems. This broadens the range of instances in which OpenVidu can be deployed, offering new opportunities and alternatives that may be more cost-effective. To deploy OpenVidu in ARM architecture, you don’t have to do anything special: just follow the default installation instructions for your desired OpenVidu deployment type. The installer will automatically detect the architecture and set up the appropriate services.
• Google Cloud Platform support for all OpenVidu deployments: we continue to expand our cloud-native deployment options. Now you can deploy OpenVidu Single Node PRO, OpenVidu Elastic and OpenVidu High Availability in Google Cloud Platform using Terraform templates.

OpenVidu can now be natively deployed in GCP

• LiveKit stack updated to v1.9.8: OpenVidu is now based on LiveKit v1.9.8, bringing all bug fixes and improvements since version v1.9.0. You can find the release notes here. Particularly relevant is the fix for issue *#3858*, which fixes a fatal problem in the connection of the services with the Redis Cluster.
• Egress updated to v1.12.0: the Egress service has been updated to v1.12.0, which includes several improvements and bug fixes when exporting media from rooms. You can find the release notes here. Particularly relevant is the change of the unhealthyShutdownWatchdogDelay value from 20 seconds to 10 minutes (see commit e86593c), preventing premature termination of egress processes under high CPU load or poor network conditions.
• Ingress updated to latest: the Ingress service has been updated to latest available commit *#e42b67a *with multiple improvements over last official release.
• Live Captions: fixed critical bug that caused slow response when transcribing 3 or more simultaneous participants in the same Room using AWS Transcribe provider (see related issue in the official LiveKit Agents repository *#3739 and PR fixing it [#4111](https://github.com/livekit/agents/pull/4111)*). Besides this, there are two new additions to the list of supported AI providers (Cartesia and Soniox) and interim transcriptions are now available for existing providers Speechmatics and Gladia.
• External MongoDB support: OpenVidu now allows configuring an external MongoDB instead of using the bundled one. You can even choose to completely disable the use of MongoDB if your use case can do without services that require a database.
• New backup and restore documentation for OpenVidu deployments: we have carefully crafted a new how-to guide explaining how to migrate your existing persistent data (recordings, analytics, monitoring data, etc.) when upgrading or changing your OpenVidu deployment. This process mainly affects the MongoDB and S3 services, responsible for persisting data in OpenVidu deployments.
• OpenVidu Angular Components updated to support Angular v20: the OpenVidu Angular Components library can now be used in applications built with Angular v20.
• Honor host time zone: OpenVidu deployments now honor their host time zone by default. Previously OpenVidu always used UTC. This proved a challenge when integrating OpenVidu with a custom app using local time zone. You can revert to UTC by providing the additional installation flag --forceUTCTimezone.
• Oracle (OCI) Single Node and Single Node PRO installation tutorials: we have created detailed step-by-step guides to help you deploy OpenVidu Single Node COMMUNITY and OpenVidu Single Node PRO in Oracle Cloud Infrastructure using its native resources.

OpenVidu Meet

Changelog

• End-to-End Encryption (E2EE) for rooms: Rooms can now enable full E2EE for audio, video, chat messages, and participant names, using a shared secret known only by participants.
• Improved meeting layout with Last N Speakers: New layout mode prioritizes the local user and up to N active speakers (1–6), improving performance and visual stability in large meetings with dozens of participants. The layout is intelligent, only updating when meaningful changes in voice activity are detected.
• Room and participant names now maintain non alphanumeric caracters: Before they were automatically removed. Now they are kept, improving internationalization and flexibility.
• Persistent data migrated from S3 to MongoDB: OpenVidu Meet data is now stored in MongoDB (recordings remain in S3), enabling better querying. Existing data is automatically migrated on upgrade.
• Improved room and recording filtering and sorting: As a result of persistent data migration to MongoDB, new status-based filters and multi-field sorting (ascending/descending) are available. Rooms and recordings are now sorted by creation date by default, instead of being ordered alphabetically by room ID.
• Fixed expired access token join issue: Logged-in users can now join meetings correctly even if a previously stored access token has expired.
• Fixed inconsistent room and recording states after server crashes: A new scheduler periodically detects and reconciles rooms or recordings left in an incorrect state due to missing webhook events.
• Fixed duplicate participant name suffixes: Repeated “_1” suffixes are no longer added when participants join with the same name.

Breaking changes

• Rooms can no longer be updated while a meeting is active: updating a room is now restricted when there is an active meeting associated with it, preventing state inconsistencies and unexpected behavior during ongoing sessions.

Stay tuned for next iterations! You can follow us on 𝕏 and a Star in GitHub is always welcome :)