Start recording on audio volume

[LaserKaspar]

Is it possible to trigger an recording when the audio for a stream goes above a certain threshold volume? I want to record my stream when it is not completely silent. I don't want it to be event based. It should just record when it hears something loud.

[inkeep-ai-assistant[bot]]

Yes, audio volume can trigger recordings in Frigate. When audio volume exceeds the configured min_volume threshold, it's treated as motion for recording purposes(1).

To enable this, you need to:

1. Enable audio detection and configure the minimum volume threshold:

audio:
  enabled: True
  min_volume: 500

(2)

The min_volume parameter sets the minimum RMS level required to run audio detection. As a guideline: 200 is high sensitivity, 500 is medium sensitivity, and 1000 is low sensitivity(1)(2).

2. Configure your recording retention mode to use motion:

When record -> retain -> mode is set to motion, any time audio volume exceeds min_volume, that recording segment will be kept(1).

3. Assign the audio role to a stream that has audio enabled:

cameras:
  front_camera:
    ffmpeg:
      inputs:
        - path: rtsp://.../sub_stream
          roles:
            - audio
            - detect

(1)

The audio levels can vary widely between camera models, so use the Audio tab in the Debug view of the Frigate UI to check the RMS levels and adjust min_volume accordingly(1).

📚 Sources:

1. Audio Detectors
2. Full Reference Config

---

Have another question? Just tag @inkeep.

[LaserKaspar]

This is my current config. I added some notes and questions to it. Hope anyone can help (;

  audio-record:
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/stream
          roles:
            - detect
            - record 
            - audio
    record:
      enabled: true
      retain:
        mode: motion   # what do I need to put here to record when audio is above a threshold?
    detect:
      enabled: false # I don't want video processing
      width: 1280
      height: 720
    audio:
      enabled: true
      min_volume: 10   # this should trigger the recording
      # can I disable audio classification so it triggers as soon as there is audio present instead of trying to classify it?

[inkeep-ai-assistant[bot]]

Your configuration has a fundamental issue: detect must be enabled for motion-based recording to work(1). Setting detect: enabled: false prevents motion detection from running, which means audio volume won't trigger recordings even though volume is treated as motion(2).

Here's the corrected configuration:

audio-record:
  ffmpeg:
    inputs:
      - path: rtsp://127.0.0.1:8554/stream
        roles:
          - detect
          - record 
          - audio
  record:
    enabled: true
    retain:
      mode: motion
  detect:
    enabled: true  # Must be enabled for motion-based recording
    width: 1280
    height: 720
  motion:
    enabled: true  # Motion detection must be enabled
  audio:
    enabled: true
    min_volume: 10

(2)(3)(1)

Regarding your specific questions:

1. "what do I need to put here to record when audio is above a threshold?" - mode: motion is correct. When audio volume exceeds min_volume, it's treated as motion and the recording segment will be kept(2).

2. "can I disable audio classification so it triggers as soon as there is audio present?" - The documentation doesn't describe a way to disable audio classification entirely while still using audio volume for recording triggers. The listen parameter defines which audio types to detect, with defaults being bark, fire_alarm, scream, speech, and yell(2)(3). However, the min_volume threshold applies before classification runs(2).

Important notes:

• Motion detection is required for object detection, so setting enabled: False for motion while leaving detect enabled will result in an error on startup(3)
• Use the Audio tab in the Debug view to monitor RMS levels and adjust min_volume appropriately(2)

📚 Sources:

1. Tuning Motion Detection
2. Audio Detectors
3. Full Reference Config

---

Have another question? Just tag @inkeep.