Notes on Talkify

Matt re­cently asked me to take a look into Talkify, which is a na­tive ma­cOS app in the same vein as Superwhisper.

The key dif­fer­en­tia­tor be­tween Talkify and its com­peti­tors is speed. Talk­ify uses the ex­ist­ing speech-to-text mod­els that Apple ships on its de­vices, which are nat­u­rally well-op­ti­mized for the hard­ware avail­able. These mod­els are fast, so Talkify is too. Ac­cord­ing to their web­site, they are as much as an or­der of mag­ni­tude faster than their slow­est com­peti­tor.

I find Talkify’s prod­uct of­fer­ing some­what con­fus­ing.

Native ma­cOS Support

The bit that I find most puz­zling here is that there is al­ready na­tive sup­port in ma­cOS for the ex­act speech-to-text model that Talkify uses.

ma­cOS even in­cludes the abil­ity to con­fig­ure hotkeys out of the box. Why would any­one use Talkify if there is a na­tive way to ac­com­plish the same task?

The UI

The ma­cOS setup is dif­fi­cult to find and it’s frankly ugly. It lacks vi­sual flair. Talk­ify fixes this. It takes that same great model and wraps it in a pret­tier UI that makes the whole ex­pe­ri­ence more sat­is­fy­ing. Is this valu­able? Yes! If a UI with a great model sucks, it makes a ton of sense to rewrap it.

Can Harper Do It Too?

Yup!

There is ab­solutely no rea­son we could­n’t take ad­van­tage of the same tech­nol­ogy Talkify uses. There is a rea­son there are lit­er­ally dozens of these text-to-speech apps. They are quite easy to build.

But Should We?

I do not think we need to add speech-to-text to Harper Desktop, or any other in­te­gra­tion. At least not yet.

The biggest rea­son is that it is a pretty big main­te­nance bur­den. If we rolled out speech-to-text as a core part of Harper Desktop, our users would ex­pect to see sim­i­lar fea­ture sets in other places. Even if adding text to speech would be easy on ma­cOS, it will likely be pretty hard to do on every plat­form that Harper sup­ports. Each plat­form might be easy, but the sheer num­ber of plat­forms makes things dif­fi­cult. In the in­ter­est of do­ing less, bet­ter, I don’t think this is an av­enue we should be ex­plor­ing yet.

On the other hand, if the Harper of­fer­ing be­gins ex­pand­ing to be more than just gram­mar and style check­ing, and in­stead to­wards a fuller suite of tools, we could to­tally in­clude high-qual­ity speech-to-text as part of that. If any­thing, I be­lieve this is a les­son in the value of a sat­is­fy­ing UI.

I am post­ing my thought process pub­licly on my blog to hope­fully find out if any ex­ist­ing Harper users would be in­ter­ested in such a fea­ture. Would the value of speech-to-text in Harper be greater than sim­ply in­stalling Talkify and Harper at the same time?