it might be one or two or three or more newlines in the last field

Well it'd be nice if your sample data would reflect that. Anyhow. I'd solve this using a simple algorithm rather than a complicated regex that backtracks god knows how often with all these .*
Consider:

use warnings;
use strict;

my @record;

while (<DATA>)
{
    if (/^\d+\~/ && @record)
    {
        print join (" ", @record), "\n";
        @record = ();
    }
    
    chomp;

    push @record, $_;
}

print join (" ", @record), "\n";

__DATA__ 
99~Markus~Holli~Mobilenum:
1234-567 , from Earth
Europe
White
Human
98~Mahesh~Babu~Mobilenum:
5678-901 , from Earth
India
Brown
Human
[download]

Output

99~Markus~Holli~Mobilenum: 1234-567 , from Earth Europe White Human
98~Mahesh~Babu~Mobilenum: 5678-901 , from Earth India Brown Human
[download]