Skip to main content

Blue Iris 6 – Tuning Guide

🔹 Overview​

This page captures practical Blue Iris tuning guidance for achieving a stable, responsive, and scalable system.

The focus is on real-world balance: image quality, reliable detection, manageable storage growth, and acceptable CPU / GPU load.

🎯 Objective​

  • Improve system stability
  • Reduce false alerts
  • Preserve useful video evidence
  • Keep storage growth predictable
  • Balance quality against performance

🧠 Key Concepts​

  • Tune one variable at a time
  • Stability matters more than maximum quality
  • Detection quality and recording quality are related, but not identical
  • AI should confirm events, not compensate for poor base setup
  • Direct-to-disk reduces system load
  • Overly aggressive settings usually create more problems than value

🛠️ Core Tuning Areas​

Stream Strategy​

  • Confirm the camera is providing the expected stream
  • Use the main stream where image detail is critical
  • Use sub-stream strategies where supported and appropriate
  • Avoid unnecessary re-encoding when direct-to-disk is the goal

Codec Choice​

  • H.264 is often the safer baseline for multi-camera stability
  • H.265 may reduce bandwidth or storage in some cases, but can increase decode complexity
  • If smoothness or stability suffers, test H.264 before making other major changes

Frame Rate​

  • Around 15 FPS is often a strong balance point for many security cameras
  • Higher frame rates increase storage and processing load
  • Match frame rate to the actual use case:
    • General surveillance: moderate FPS
    • Fast motion or identification zones: higher only where justified

Resolution​

  • Use enough resolution to support the actual objective
  • Higher resolution does not always improve usable evidence if the scene is too wide
  • Camera placement and field of view often matter more than simply increasing megapixels

Recording Mode​

  • Decide whether the camera should record continuously, on trigger, or on a hybrid approach
  • Direct-to-disk is preferred when supported
  • Confirm pre-trigger and post-trigger behavior matches the real event length you want to capture

Motion Detection​

  • Start with conservative motion settings
  • Define zones carefully
  • Adjust object size and contrast thresholds to reduce noise
  • Tune for the actual scene:
    • trees and shadows
    • rain and insects
    • water movement
    • headlights and reflections

AI Confirmation​

  • Use AI to improve event quality, not as a substitute for poor motion tuning
  • Begin with moderate confidence settings
  • Confirm the AI is seeing a good image source
  • Review confirmed vs unconfirmed events before enabling aggressive alerting

Alerts​

  • Do not enable every alert action at the beginning
  • Confirm trigger reliability first
  • Then confirm AI behavior
  • Then enable notifications
  • Keep alert workflows simple until the system proves stable

Storage​

  • Confirm New, Stored, and Alerts locations are intentional
  • Check retention rules and size caps
  • Watch how fast storage grows after changes to FPS, resolution, and recording mode
  • Make storage decisions based on actual event volume, not assumptions

Performance​

  • Watch CPU, memory, storage responsiveness, and GPU behavior together
  • Hardware acceleration can help, but only if it is stable in your environment
  • Too many simultaneous high-resolution streams can degrade the whole system
  • Add load gradually and validate after each change

📊 Recommended Baseline​

Stable Starting Point​

  • Codec: H.264
  • Frame rate: ~15 FPS
  • Recording: direct-to-disk where appropriate
  • Motion: conservative tuning
  • AI: confirmation after motion is reasonably clean
  • Alerts: basic only until validated
  • Storage: confirm retention before scaling camera count

🧠 Real-World Notes​

  • H.264 is often more forgiving than H.265 in larger mixed-camera systems
  • Pre-trigger is one of the easiest settings to underestimate
  • False alerts usually come from scene conditions and trigger design, not just “bad AI”
  • Wide scenes often benefit more from better placement or tighter field of view than more pixels
  • A clean, boring setup usually outperforms an over-tuned setup over time

⚠️ Common Mistakes​

  • Turning on too many advanced features at once
  • Increasing frame rate without a real need
  • Using maximum quality everywhere
  • Enabling alerts before validating detection quality
  • Ignoring storage growth until after the system is already busy
  • Changing multiple variables at once and losing the ability to isolate the cause

📊 Related Pages​

✅ Result​

This page provides a practical tuning framework for keeping Blue Iris stable, useful, and scalable in real deployments.