The man page for RTF::Writer doesn't discuss this, but what if you want to create an rtf file containing unicode text in some obscure language that uses a wacko writing system? I played with a sample unicode file (in Thai), managed to load it into MS Word and "Save as ... RTF"; this showed me the rtf expression for unicode characters. Well, all we have to do is express our non-ascii unicode characters that way, pass the result to RTF::Writer, and we're done...

#!/usr/bin/perl

use strict;
use RTF::Writer;

die "Usage: $0  file.txt\n  (this will create file.rtf)\n"
    unless ( @ARGV and $ARGV[0] =~ /\.txt$/ and -f $ARGV[0] );

# input file is expected to be utf8

open( I, "<:utf8", $ARGV[0] ) or die "$ARGV[0]: $!";
my $utf = do { local $/; <I> };  # slurp it

# here's the magic part: replace each wide character with
# "\uN\5f", where "N" is the decimal numeric codepoint:

$utf =~ s/([^[:ascii:]])/sprintf("\\u%d\\'5f",ord($1))/eg;

( my $out = $ARGV[0] ) =~ s/txt$/rtf/;

my $rtf = RTF::Writer->new_to_file( $out );

my @pars = split( /\n+/, $utf );
$rtf->prolog( title => $out );
for my $par ( @pars ) {
    $rtf->paragraph( \$par );  # need to pass $par by reference
}
$rtf->close;
[download]

Is that easy, or what?

Update: Caveat emptor: YMMV, especially depending on what language you're dealing with (I've only tried Thai so far). I don't know how/whether RTF handles bidirectional text (Arabic- or Hebrew-based writing systems), and various Indic scripts (Hindi, Tamil, etc) might have weird problems if the rtf recipient doesn't know the "special rules" for rendering those languages. But the "normal, tidy, linear" languages (Cyrillic, Greek, Chinese, Korean, etc) should be fine (assuming you have the appropriate unicode fonts available).