Use case
I use Hypherion/HypherHDR (a fork of Hypherion) with a custom HACS Hyper-HA integration to get the average color of my TV screen.
Now, the aim is to do a poor mans ambilight with just a smart light that shines somewhat the correct light.
Of course, this is more of a fun project and to test out whether that is actually useful or how it works. Of course, I totally see and would recommend to use a proper LED stripe and directly let it control via that. But well… here we go for just a ambient room I may say. 
But well… maybe with the right tuning it turns out to be useful. 
Approach
The integration exposes the “average screen color” already as a HEX value:
I guess that is most useful to use here.
Code so far
With some AI help and trying out stuff, I was able to generate some working automation:
alias: Simple ambilight
description: ''
triggers:
- trigger: state
entity_id:
- sensor.yourdevice_hyperhdr_averagecolor
for:
hours: 0
minutes: 0
seconds: 0.5 # specifically throttled, as I noticed fast enough light changes are unrealistic anyway, so "smoother" values are better
conditions:
- condition: state
entity_id: light.yourlight
state:
- 'on'
actions:
- action: light.turn_on
data:
transition: 0.25 # again for smoothness, but should still ensure fast updates
rgb_color: >
{% set color =
states('sensor.yourdevice_hyperhdr_averagecolor')
| replace('#', '') %}
{{ [
color[0:2] | int(16),
color[2:4] | int(16),
color[4:6] | int(16)
] }}
target:
entity_id: light.yourlight
mode: restart # useful to prevent stale color updates
This works surprisingly well already, though one can often feel it “lags behind” the TV image. As indicated above, I already added some “smoothing”. To properly do that over longer time spans without wild color changes, “smoothing” over a longer time span would be needed. Aka, mathematically speaking, calculating the average of the color over time?
In my imagination, this would capture the “scene” over a longer period of time, maybe even dunno, the last, 10 seconds or so? Then a nature documentary in a jungle would be primarily green while something else would be different.
As said, it’s kind of an experiment.
Question
Now obviously Home Assistant does have an average function: That said, again obviously, calculating an average of a color obviously is harder, but possible. Apparently, depending on the color space/representation you use, there are also a lot of different algorithms.
Also cross-posted to the HA integration discussions forum I use to get their ideas/opinions on that.
