Skip to content
NovaVMS Manual

Choose which stream feeds detection and which feeds recording

Choose which stream feeds detection and which feeds recording

Use this procedure when you want to tune a camera’s cost and quality — for example, run AI on a low-resolution sub stream to keep LLM costs down while recording event clips from the high-resolution main stream.

Procedure

  1. Go to Cameras and open the camera you want to tune.
  2. Click the Settings gear icon on the camera detail page.
  3. In the settings panel, scroll to the Stream section.
  4. Set Detection stream to sub or main.
  5. Set Recording stream to main or sub.
  6. Click Save.

Changes apply to the next event. Events already recorded keep their original stream assignment.

Pick the right assignment

StreamPurposeRecommended codecRecommended resolution
SubDetection (AI analysis)H.264480p – 720p
MainRecording (event clips)H.265 or H.2641080p – 4K

The tradeoff:

  • Low-res detection means the VLM processes smaller frames. Lower per-event LLM cost, faster analysis, slightly lower tag precision on small objects.
  • High-res recording means event clips are large. Bigger storage footprint per event but usable footage for evidence export.

If a camera does not expose a sub stream, set both fields to main.

Common variations

  • Detection only, no recording: set Recording stream to sub. Clips will be small and lower quality. Useful for cost-sensitive cameras where the AI tags matter more than the footage.
  • High-fidelity detection: set Detection stream to main when you are running complex prompt packs that benefit from extra detail (licence plates, small objects).

If this didn’t work

  • If the dropdown only shows one option, the camera advertises a single ONVIF profile. You cannot split streams on that camera.
  • If event clips look wrong after a change, open the event — the stream used is recorded on the event itself, not inferred from the current setting.