Release Notes

The release notes track changes to Wowza's Video Intelligence Framework (VIF) and its two components: the Video Intelligence Controller (VIC), the plugin inside Wowza Streaming Engine, and the Video Intelligence Service (VIS), the GPU-accelerated inference service. To learn more about how Wowza VIF works, see About Wowza Video Intelligence Framework. Releases are listed below, newest first.

Changelog

September 7, 2026

VIC 1.1.0 and VIS 1.1.0

Important! VIF API v2 is now standard, and v1 is deprecated. The v2 REST API is now the supported interface, and the Manager UI runs entirely on v2.

• Added on-demand (VOD) analysis, so you can run any detector on uploaded video files. A new VOD area in the Manager lets you upload files, submit and monitor jobs, filter the job list by state, and view results, result files, and thumbnails. The VOD application is configured by default, so it's ready to use without extra setup. For more information, see Analyze VOD files with VIF.
• Introduced the v2 REST API and moved the Manager UI onto it. You can now tune a running stream's configuration live and apply per-stream configuration overrides. For more information, see the Wowza Video Intelligence Framework v2 API. 
• Redesigned the Manager UI across the dashboard, stream configuration, default configuration, and playback pages. See Configure VIF in Engine Manager for more information.
• 
  Improved VLM analysis.

August 13, 2026

VIC 1.0.3

• Added NVIDIA Nemotron Nano 12B VL, Google Gemma 3 4B, NVIDIA Cosmos3 Edge, and NVIDIA Cosmos3 Nano as selectable VLM models. Selecting a model alone does not deploy it. You still need the matching VLM sidecar running on the VIS server.
• Renamed the VLM "Endpoint URL" field to "VLM Server" and made it accept a plain address such as vlm.docker:8000; VIC adds the http:// scheme and /v1 path on save, so localhost:8000 is stored as http://localhost:8000/v1.
• Added a Verify button that checks the VLM server is reachable and reports which model it's serving. If the server serves a single model that differs from the one you selected, Verify switches your selection to match it so you can save it.
• Fixed VLM streams failing to start when no endpoint or model was configured anywhere; VIC now falls back to the bundled vLLM defaults (http://vlm.docker:8000/v1) and logs a warning so you know the stream is using the built-in endpoint rather than your own.

August 12, 2026

VIC 1.0.2

• Configured event listeners now appear in a single list, so you can see everything you've set up at a glance.
• Removed alert-style coloring from CPU/GPU statistics that could imply a problem when none existed.
• Tooltip help now displays correctly.
• The maximum tokens setting is now limited to a valid range.
• Removed an unnecessary warning that could appear when the keyframe interval was fine.
• Thumbnail selections now stay in place as you navigate.
• Added error handling to the HLS playback page when streams are no longer available.