Audio spectrum display (ESP32, WS2812B)

and foils from a broken LCD TV as a screen? there are several layers of different foils and one with 4mm plexiglass and everything is the size of the matrix, i.e. large sheets, and a broken TV can be found in a dumpster or probably in any normal RTV service you can probably get such foils for free I have a lot of it, because I leave plexiglass from broken matrices and I give foils to my daughter and she paints on it with markers

I made a preliminary inspection of the laptop display elements. That white sheet absorbs too much light, but that bottom silver one is interesting. You have to do trials

Well, actually, I`ll ask. I`m struggling with a certain VFD display, with a lot of cooperation with AI (I`m a rather poor programmer). Your the FFT implementation using the Arduino library is very cool. For me, even with the help of AI, it displays nonsense. Do you have an idea for a function using ArduinoFFT to return data as an array (f, a), where f is the frequency of the "bar" and f is the amplitude?
Because for now it shows nonsense.

At the beginning of the article @KJ mentioned that there is a lack of fluidity in the movement of the posts. Rightly so, by the way. I took the plunge and decided to rework the whole unit.
First of all, I doubled the vertical resolution of the bar from 16 to 32 LEDs in two columns, which increased the number from 32 to 64 LEDs per bar. As a result, the total number of LEDs has increased from 640 to 896 (40%). The typical LED strips were no longer suitable. I had to make a PCB for the LEDs. And, of course, new fenders and hoods.

.



The original version only pulled 20 FPS. After adding the LED, it was already a disaster and not FPS.
We had to get on with writing new software for the ESP32.
The result was an acceleration to 37 FPS. Practically the frames were limited by the amount of WS2812B. For such a quantity, faster is not possible.
FPS = 1000/quantity_WS * 0.03.
For 896 LEDs, FPS = 37.
It could stay, because the smoothness has improved considerably. But.... I decided to go further .
The final panel is controlled from two ESP32s. ( At the cost of the whole panel, 12PLN makes no difference). Each operates half of the panel. Both cores and a timer are harnessed in each.
The number of frames now is 72. I have also introduced audio sample averaging and a small delay in the descent of the bars. Scaling, as suggested by commenters, is now logarithmic.
As a reminder, the initial version.
.
New version. In the video, 'average' audio sampling and bar descent settings. Can be sped up or slowed down.
.