Hacker News Reader

Comment by wewewedxfgdf | Hacker News Reader

say is only 193K on MacOS

  ls -lah /usr/bin/say
  -rwxr-xr-x  1 root  wheel   193K 15 Nov  2024 /usr/bin/say

Usage:

  M1-Mac-mini ~ % say "hello world this is the kitten TTS model speaking"

View on HN

That’s not a far comparison. Say just calls the speech synthesis APIs that have been around since at least Mac OS 8.

That being said, the ‘classical’ (pre-AI) speech synthesisers are much smaller than kitten, so you’re not wrong per se, just for the wrong reason.

The linked repository at the top-level here has several gigabytes of dependencies, too.

SAM on Commodore 64 was only 6K:

https://project64.c64.org/Software/SAM10.TXT

Obviously it's not fair to compare these with ML models.

And what dynamic libraries s it linked to? And what other data are they pulling in?

`say` sounds terrible compared to modern neural network based text to speech engines.

Sounds about the same as Kitten TTS.

To me it sounds worse, especially on the construction of certain more complex sentences or words.

Tried that on 26 beta, and the default voice sounds a lot smoother than it used it.

Running `man say` reveals that "this tool uses the Speech Synthesis manager", so I'm guessing the Apple Intelligence stuff is kicking in.

Nothing to do with Apple Intelligence. The speech synthesiser manager (the term manager was used for OS components in Classic Mac OS) has been around since the mid 90s or so. The change you’re hearing is probably a new/modified default voice.

> Nothing to do with Apple Intelligence Sorry, the clarify - I mean in macOS beta 26, the `say` command's output matches the settings set in the System Settings -> Apple Intelligence & Siri -> Voice > Select... menu.

> speech synthesiser manager (the term manager was used for OS components in Classic Mac OS)

Especially fun to play with on the rainbow iMacs back then, too.

If you make a shell script that calls say, that script will be even smaller!

Reply on news.ycombinator.com