It was several years in the past, but I have used Template with unicode a lot. You can have UTF-8 encoded templates and UTF-8 strings in variables - you just need to declare them consistently.

The very first configuration parameter documented in Template::Manual::Config is ENCODING. I'll quote it here because it is so short:

ENCODING

The ENCODING option specifies the template files' character encoding:

my $template = Template->new({
    ENCODING => 'utf8',
});
[download]

A template which starts with a Unicode byte order mark (BOM) will have its encoding detected automatically.

So, the following program works as I'd expect:

use Template;

use open ":std", ":encoding(UTF-8)";

my $dogface = "\N{DOG FACE}";

my $template = <<"ENDOFTEMPLATE";
Dog face from template: $dogface
Dog face from variable: [% dogface %]
ENDOFTEMPLATE

my $tt = Template->new(ENCODING => 'UTF-8');
$tt->process(\$template,{dogface => $dogface});
[download]

The automatic BOM detection is a handy band-aid if you have a mixture of Latin1 and UTF-8 encoded templates and don't want to re-code them.