Looking at the datasheet page 3 we see that the minimal voltage for a high level input is 0.7*Vdd. Vdd is the supply voltage of the LED, so 5V in this case. 0.7 * 5V = 3.5V
Data level output of ESP
The ESP can only output 3.3V.
Conclusion
The output of the ESP is in theory not high enough to drive the WS2812B LEDs.
A levelshifter does fix that issue by changing the datalevel from 3.3V to 5V.
But it works without one in Video XY
As said above: " …in theory…". Values from the Datasheet and the reality are two different things.
The values in the datasheet mean that it can garantee a good read of a HIGH level at 3.5V or more. It can work below 3.5V, but there is a chance does not get a good read.
And now?
If you want a reliable setup, then integrate a recommended levelshifter. Check the /shifter command for those.
This is also a good video from VCS from years ago. They only really talk about 5v data (levelshifted), but it is still a good breakdown of what happens to data signals. https://www.youtube.com/live/9ttEAyiyTAI