I would suggest getting rid of non-text characters from inputs. Control characters and other non-printable, strange characters do not really make sense in this problem, and are going to cause pain with presentation across languages (like it happened in the Go translation and was shrugged off for some reason as irrelevant). Limiting the alphabet to text characters would significantly improve user experience across all languages.

Unless the emojification is meant as a textual encoding of binary data (like Base64), then, well, I guess any byte is valid. But in such case, translations should take special care for presenting non-printable characters in inputs and outputs.

Duplicate