I'm running Sierra (10.12.5) and I've been using the text-to-speech feature to read news articles. The system stumbled over the phrase "defeat ISIS" today, saying DEH-feh-CHURring EYE-sis. The system can pronounce "defeat" and "ISIS" correctly as separate words, but if I put them together, "defeat" gets the incorrect pronunciation.

While probing this oddity further I've discovered that:

  1. Of the English speaking voices, only the Samantha voice has this flaw.
  2. The DEH-feh-CHURring pronunciation occurs if you follow "defeat" with any word at all. Even random strings produce the error.

Is this user-fixable short of giving up on the Samantha voice?

  • Hmm…it seems as though it interprets "defeat" as being the (nonexistent?) word "defeaturing." (Try having it say "defeat me.")
    – aaplmath
    Jul 5, 2017 at 17:52
  • Fascinating. It doesn't even say the "me" part.
    – Kyle Jones
    Jul 5, 2017 at 17:58
  • The issue appears to steam from the fact that Samantha will interpret "feat" as being an abbreviation of "featuring" when it is followed by any other word. However, this is poorly implemented ("feat me" has the same problem as "defeat me") and it appears that "defeat" is still interpreted as somehow containing the "feat" abbreviation. The easiest work-around is to follow "feat" or "defeat" with a comma, though that might be difficult if you're talking about something as large as an entire article.
    – aaplmath
    Jul 5, 2017 at 18:05
  • @aaplmath Thanks for that hint. Appending a dash to "defeat" seems to work also and doesn't make the reader pause, which is nice. This won't help with the articles I have the web browser read to me but there's other text that I feed to the "say" command that I can now munge to correct the pronunciation.
    – Kyle Jones
    Jul 6, 2017 at 3:12

1 Answer 1


Speech Synthesis in OS X

As a user, you can affect macOS's text to speech pronunciation and approach. Everything you need to know is in Apple's Speech Synthesis in OS X.

You can provide TUNE Format information to control how the speech is generated. An Apple developer tool called Repeat After Me may also be useful in this endeavour.

Also see add pronunciation to text to speech.

Let Apple Know

However, you have encountered an interesting and serious bug in Apple's text to voice system. I recommend telling Apple's engineers directly either through the macOS feedback form or via the Bug Reporter service.

The problem can be quickly demonstrated with these two Terminal commands:

say -v Samantha "defeat ISIS"


say -v Alex "defeat ISIS"
  • Thanks for the links. None of it looks like a user-oriented solution, though there is fine-grained control for developers writing apps.
    – Kyle Jones
    Jul 6, 2017 at 3:21

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .