Wireless audio is not sent as-is. It is compressed, packaged, transmitted, buffered and then decoded, and every stage adds a few milliseconds. Added together, the delay can reach a couple of tenths of a second, which is enough for lips and speech to visibly disagree.
The buffer is deliberate. Without it, every small radio interruption would produce a click or a dropout, so the receiving device keeps a short reserve of audio in hand. Larger buffers mean smoother sound and more delay; smaller buffers mean the opposite.
Video apps often compensate by delaying the picture to match. That is why films look fine but games and live instruments feel wrong: an app that knows the audio delay can offset it, while anything reacting in real time cannot.
Different compression methods have very different delays, and both sides must support the same one. A headset capable of low-latency audio falls back to a slower method when the phone or laptop does not support it.