.MP3 file format which worked for my XVR/DVR (Dahua re-branded):
Length: 24 seconds
MPEG-1 layer 3
160 kbps, 926 frames
44100 Hz Joint Stereo
The voice recorded has to be longer than a few seconds (if you have 3 seconds, you will hear only a short noise)
I used an online text to speech converter to create the voice file (I've wrote as text: network disconnected network disconnected network disconnected ..... for more than 10 times => a .wav file which plays more than 10 times "network disconnection")
I used an online .wav to .mp3 converter with the settings of the format mentioned above.
The DVR was able to play this .mp3 file.
I also found that the DVR plays the .mp3 file once even though the error still occurs, the buzzer sounds but the voice is played only once. So if you want 5 minutes of voice then create a 5 minutes voice file. (I tested only 24 seconds)
I attached a .rar file with the .mp3 sample which I've tested.
good luck
speech.rar