home-assistant.io/source/_components/binary_sensor.ffmpeg.markdown

layout, title, description, date, sidebar, comments, sharing, footer, logo, ha_category, ha_release

layout	title	description	date	sidebar	comments	sharing	footer	logo	ha_category	ha_release
page	FFmpeg Binary Sensor	Instructions how to integrate a varius ffmpeg based binary sensor	2016-08-25 08:00	true	false	true	true	ffmpeg.png	Binary Sensor	0.27

The ffmpeg platform allows you to use every video or audio feed with FFmpeg as varius sensors in Home Assistant. Avilable are: noise, motion. If ffmpeg process is brocken, sensor going to unavilable. It exists a service to restart a instance with binary_sensor.ffmpeg_restart.

You need a ffmpeg binary in your system path. On debain 8 you can install it from backports. If you want HW support on raspberry you need self build from source. Windows binary are avilable on ffmpeg homepage.

{% linkable_title Noise %}

To enable your FFmpeg with noise detection in your installation, add the following to your configuration.yaml file:

# Example configuration.yaml entry
camera:
  - platform: ffmpeg
    tool: noise
    input: FFMPEG_SUPPORTED_INPUT
    name: FFmpeg Noise
    ffmpeg_bin: /usr/bin/ffmpeg
    peak: -30
    duration: 1
    reset: 20

Configuration variables:

• input (Required): A ffmpeg compatible input file, stream or feed.
• tool (Required): is fix set to 'noise'.
• name (Optional): This parameter allows you to override the name of your camera.
• ffmpeg_bin (Optional): Default 'ffmpeg'.
• peak (Optional): Default -30. A peak of dB to detect it as noise. 0 is very loud and -100 is low.
• duration (Optional): Default 1 seconds. How long need the noise over the peak to trigger the state.
• reset (Optional): Default 20 seconds. The time to reset the state after none new noise is over the peak.
• extra_arguments (Optional): Extra option they will pass to ffmpeg. i.e. audio frequence filtering.
• output (Optional): Allow you to send the audio output of this sensor to a icecast server or other ffmpeg supported output. i.e. to stream with sonos after state is trigger.

For playing with values:

ffmpeg -i YOUR_INPUT -vn -filter:a silencedetect=n=-30dB:d=1 -f null -

{% linkable_title Motion %}

FFmpeg don't have a motion detection filter so it use a scene filter to detect a new scene/motion. In fact you can set how big a object or size of image they need change to detect a motion. The option 'changes' is the percent value of change between frames. You can add a denoice filter to video if you want a realy small value for 'changes'.

To enable your FFmpeg with motion detection in your installation, add the following to your configuration.yaml file:

# Example configuration.yaml entry
camera:
  - platform: ffmpeg
    tool: motion
    input: FFMPEG_SUPPORTED_INPUT
    name: FFmpeg Motion
    ffmpeg_bin: /usr/bin/ffmpeg
    changes: 10
    reset: 20
    # group feature / default not in use
    repeat: 0
    repeat_time: 0
    

Configuration variables:

• input (Required): A ffmpeg compatible input file, stream or feed.
• tool (Required): is fix set to 'motion'.
• name (Optional): This parameter allows you to override the name of your camera.
• ffmpeg_bin (Optional): Default 'ffmpeg'.
• changes (Optional): Default 10 percent. A lower value is more sensitive. I use 4 / 3.5 on my cameras. It describe how mutch of two frames need to change to detect it as motion. See on descripton.
• reset (Optional): Default 20 seconds. The time to reset the state after none new motion is detect.
• repeat (Optional): Default 0 repeats (deactivate). How many motion need to detect in repeat_time to trigger a motion.
• repeat_time (Optional): Default 0 seconds (deactivate). The time to repeats befor it trigger a motion.
• extra_arguments (Optional): Extra option they will pass to ffmpeg. i.e. video denoise filtering.

For playing with values (changes/100 is the scene value on ffmpeg):

ffmpeg -i YOUR_INPUT -an -filter:v select=gt(scene\,0.1) -f framemd5 -