[ANDROIDHASSAN]

punc_norm in mtl_tts.py appends . unless the text already ends in one of its sentence enders. Hindi and Arabic are supported languages, but their sentence punctuation was not in that set, so their input got a second stop:

before                         after
'मेरा नाम हसन है।'  -> 'मेरा नाम हसन है।.'    'मेरा नाम हसन है।'
'श्लोक ॥'          -> 'श्लोक ॥.'             'श्लोक ॥'
'كيف حالك؟'        -> 'كيف حالك؟.'           'كيف حالك؟'
'مرحبا،'           -> 'مرحبا،.'              'مرحبا،'
'Hello there'      -> 'Hello there.'         'Hello there.'   (unchanged)
'नमस्ते'           -> 'नमस्ते.'              'नमस्ते.'        (unchanged)

This adds the danda ।, double danda ॥, Arabic question mark ؟ and Arabic comma ، to sentence_enders, next to the CJK marks already there (the ASCII comma is already an ender, so the Arabic comma now behaves the same). All four characters are in the multilingual tokenizer's vocabulary, so the text now ends on the real punctuation token instead of an extra ..

Checked by running punc_norm from this branch on the inputs above.