macOS Voices and Touchpads
I often make use of voice read-aloud TTS functionality. For instance in Claude you can click “Read Aloud” at the bottom of a response and have it read the text. This can be useful when doing other things that don’t require complete focus. For instance cooking or chores.
The thing is, though, that the Claude desktop app (at least up to what I have today on macOS, v1.24012.9) has a subpar text-to-speech engine. It is technically impressive, but adds bizarre “stoner” pauses throughout the reading. It also adds accentuations on certain words during reading, not unlike a kid giving a class presentation of material they didn’t prep for at all. It is very strange, and makes it almost impossible to mentally parse, at least for me. You have to fully focus and then reassemble the sequence of words to re-interpret in your mind.
Of course macOS itself has a text-to-speech engine. It actually has multiple such systems, disjointed and isolated (for instance install a voice in Read & Speak, and it won’t be installed in VoiceOver, as they seem to just duplicate the underlying functions as some system group just copied and pasted a whole subsystem), but the primary one for this purpose can be found in Accessibility / Read & Speak. There you can configure a voice and a hotkey to read aloud the selected section, for instance Option+ESC.
It works great, though many of the system voices are terrible (click the little ⓘ to the right of the system voice to see other options), having been dragged forward through the years without deprecation. Early industry “Would you like to play a game” voices like Bruce are still in there. And it’s all just provided in a list with no best-before date or quality/generation indicators, though some have a preview function.
One of the voices in there - Alex - is 885MB. You would think it’s probably a fairly modern voice, right? I installed it to give it a try.
It’s terrible. I mean, terrible is overly critical, and when it was released with Mac OS X Leopard back in 2007, it was cutting edge. But again, the list of voices has zero indication of the age or modernity. 95% are completely obsolete options, and I guess it relies upon a historic knowledge of the Apple platform to know which are from which generation.
So I went to delete it – why waste almost a GB for this – to find there is zero affordances for delete anywhere in the voices control. No right click, delete key, special sequence. There are no command line options to purge a voice. There are tutorials online detailing where to find the voice downloads on the file system to manually purge, but none of them were accurate for macOS 27, and apparently these voice files move around with almost every iteration.
It turns out that to delete the voice you need to use a touch surface and do the classic iOS swipe left gesture over the row, and that is your only option. This is a side effect of macOS adopting the iPad style of system settings that happened back in Ventura. There is zero indication of this anywhere.
While I have a magic mouse, I usually keep it tucked away in my desk, using a Razer gaming mouse for my primary instead. The magic mouse surface tracking isn’t super great – it is a terrible mouse for precision work – and the touch surface would constantly lead to issues in apps like Affinity where I would constantly be doing unwanted zooms in and out, pans, and so on. And while I can disable the touch surface or specific gestures like zoom, with it disabled it is just a mediocre mouse and loses the one benefit to justify all the other problems.
I pulled the Magic Mouse out, charged it up, and connected it for the sole purpose of swiping left over a row in the system settings. Once that was accomplished it went back into hibernation in the drawer.
In the end I went with Siri “Voice 4”. Great all around voice.
WWDC and Pre-Recorded Presentations
WWDC happened almost two months ago. This is a bit delayed, but I meant to comment on the way they present the keynote, state of the platform, and various other elements of the event.
I love that they spend the time polishing everything. That they put in the effort to maximize the information per unit of time, cutting out most of the slop and filler that can slow such a presentation down with irrelevant errata and so on1.
They moved to this format during COVID, and so far they’ve stuck with it. I hope they continue. There is no value in seeing people fumble over delivery or suffer the yips, wifi not connecting, people fighting with a deck remote, the time sucking transition between presenters, random protestors for some world happening, and so on. Add that it is more accommodating, accessibly including presenters who wouldn’t be able to take part if it was a “one shot and you better get it right in front of a big crowd” format.
I mention this because each WWDC, tech venues fill with critics denouncing it. On HN, an absolutely miserable cesspool of “who moved my cheese?” resist-all-change angry sorts2, it’s free karma to be first to post something like “They try so hard to do a polished presentation that everything is kinda fake and unauthentic. I don’t understand how this attitude survived so many years”, then universally declaring that “No one likes this”.
I like it. If you’re just delivering a torrent of product updates, pre-record it and polish it to perfection.
Maybe I’m in the minority. I cannot stand those podcasts where people just sit around and riff about random bullshit, for instance. It feels incredibly lazy and like a massive waste of everyone’s time. It’s just slop. But a lot of people like that.
Several commentators then reminisce about the good old days of WWDC with a live audience3. About the “organic” response that told them how to respond to the presentation. Similar to how the laugh track in the Big Bang Theory tells them if something was “funny”4.
Are they aware that Apple, and almost all companies hosting their own show, stuffed the audience with employees? Those “organic” hoots and hollars about random things were the laugh tracks of the tech world. People were really fooled by this?
Footnotes
-
Though they often get a “focus” that can hijack the event. I remember one year when they spent about 80% of the keynote talking about how recyclable their boxes and products were, over and over and over again. ↩
-
To do the “who moved my cheese”5 thing, HN is kind of terrible in 2026, and has fallen so far from the greatness it once held. The moderation is good and the site is fast and responsive, but the average user has declined precipitously.
HN used to be a leading-indicator of rising techologies, languages, approaches, and so on. It was where you would find the next thing years in advance and kind of ride an early wave. Now it’s a bunch of tired, stale people lamenting the state of things, reminiscing about a time that never was, and doing everything possible to resist not only the future, but even the present. It’s just sad. There are still good signals, but they’re hidden among a lot of garbage. ↩
-
Though somehow people always cite the event where Steve Jobs was told by an audience member that he doesn’t know what he’s talking about. Only that event was explicitly a Q&A session – which generally needs a live audience – and was held as a “fireside chat”. It has zero relevance to a keynote or similar “we deliver the information” situation.
Yes, if you’re doing a Q&A, do it live.
Humorously the top comment on that YouTube video – with 10k+ upvotes – is just entirely wrong6. The audience member was a consultant at Sun Microsystems. Sun was one of the core contributors to OpenDoc, which Jobs had declared that they were killing off, and his point was that while Jobs was embracing Java, also from Sun, he didn’t believe Jobs understood the different realms that OpenDoc and Java occupy, and that they worked best together. ↩
-
As much as people rail about AI slop, just watch that BBT video and marvel in the absolute garbage that not only gets aired on a major network, it somehow rose to be the #1 most viewed show. We don’t need AI to make slop. ↩
-
The book “Who Moved My Cheese” is horrible. It’s a garbage, worthless read. But everyone understands the meaning of the title, so it at least has that. ↩
See Also
- On AI Criticism
False dichotomies muddy the waters