Spatial computing has a keyboard problem

But there's a solution literally in front of you.

One month after Vision Pros started landing on doormats across America, two early use cases are emerging for this newfangled spatial computing thing.

The first is content consumption. Headsets are great for watching movies because of the unparalleled immersion they provide from the comfort of your home or airplane seat. There’s a rough consensus that this is a use case the Vision Pro already does very, very well.

Image credit: Apple Inc.

The second is getting work done. People are hoping the headset can make them more productive. But this use case isn’t completely ready for prime time. From lack of software, to battery life, to comfort, there are many reasons why we aren’t living Apple’s spatial computing utopia yet.

Image credit: Apple Inc.

What would it take to get us there?

Well, one of things we need to solve is typing.

Keyboards are flat, but headsets are not

Apple and Meta are both trying to position spatial computing as the next major computing paradigm. For this to succeed, people need to be able to use headsets to do real work. To do real work, you need a real keyboard.

Unfortunately, headsets and keyboards literally don’t go together: keyboards are flat, but headsets aren’t. Geometry makes it impossible to neatly couple them. The two really want to be separate objects.

Image credit: Apple Inc.

Consequently, aspiring spatial computing devotees can’t count on always having a keyboard with their headset. Even if you invest in a dedicated keyboard, it’s easy to forget to bring it along, because there’s no elegant way to physical attach it to the headset.

(This is not the case for other portable computing devices, because their displays are flat. This flatness makes it possible to mate them with a keyboard, or to display one onscreen.)

Real work needs real keyboards

History strongly suggests that general-purpose computing requires a physical keyboard.

Looking at the broad arc of the two foundational computer input methods—pointing and typing—it’s notable that one is becoming more ephemeral, whereas the other is not.

Pointers have gone from chunky wired mice with two buttons → to svelte wireless mice with once button → to smooth trackpads that are all button → to the Vision Pro's nearly-telepathic eye tracking.

Keyboards, on the other hand, did get thinner and sleeker over the decades. But then they hit a plateau, and attempts at taking them past a certain level of immateriality (butterfly keys, touch bars) have been met with loud user resistance.

If anything, keyboards are rematerializing. Observe the return of physical function keys in the MacBook series, or the increasing popularity of premium mechanical keyboards.

Image credit: Electronic Materials Office, Clicks Technology

Typing defies ephemeralization

Neither virtual keyboards nor speech-to-text are substitutes for the real thing.

Once the novelty of typing on a levitating virtual keyboard wears off (which takes about three seconds), using a virtual keyboard to write anything longer than monosyllabic words makes you want to throw something heavy out a window.

Even Apple's implementation is frustrating to use, which means virtual keyboards aren’t something a couple billion more billion dollars can easily solve.

Image credit: Apple Inc.

There have been some interesting experiments out of Meta's Reality Labs around virtual keyboards which are mapped to a physical surface, like a desk. The fact they are boasting such impressive WPM numbers but haven’t shipped it yet suggests there’s some fatal flaw to this approach.

Image credit: Meta Platforms, Inc.

AI speech-to-text models like Whisper have gotten stunningly good. We use Whisper 3 to power Softspace’s speech-to-text button. But speech-to-text is only really good for the initial inputting of text. It’s not good for fine editing of text, which is just as important when doing real work.

Also, there are just too many moments in one’s day when speaking the text you want to input, out loud, would be awkward or inconsiderate or otherwise undesirable.

There’s a solution in front of you

If real work needs a keyboard, but headsets can’t come with one, is general-purpose spatial computing doomed?

Not if spatial computing doesn’t completely displace legacy computing devices in the near-term, but complements them instead.

Early reports indicate that people really love using their Vision Pro as an external display for their MacBooks. The most obvious advantage to this setup, compared to using a visionOS app, is that all your programs and data are already there.

Image Credit: Apple Inc.

But an under-appreciated reason why this feels so good is that you’re using an industry-leading trackpad and keyboard with the headset.

Apple has taken the logical next step of allowing your laptop’s keyboard to input text not only within the virtual display of the MacBook, but also in native visionOS apps.

Unfortunately, Meta’s Quest doesn’t have the system-level ability to do this. Users who want to use a physical keyboard with their Quest have to pair a dedicated Bluetooth keyboard.

Unless they’re using Softspace.

We built a web keyboard for Softspace

Using Softspace with a physical keyboard is a far superior experience to using it without. Even if you’re browsing the web and not writing lot of text, this is true.

But we know from the metrics that only a small minority of users ever use the app with a Bluetooth keyboard. We understand why: dedicating a keyboard to the headset is a big hassle, especially if it’s only for a single app.

So we built a way for users to use a keyboard on any other (internet-connected) device inside Softspace, via a web app that conveys keystrokes to the Quest.

Here’s how it works: in Softspace, you generate a keyboard pairing code. On your laptop (or tablet, or even phone), you go to keyboard.soft.space and enter this code. For additional security, a second confirmation code is displayed in the headset, which you input into the keyboard web app.

Once that’s done, you can just start typing in Softspace using your laptop.

If the headset and laptop are on the same local network, the latency is slightly noticeable but not problematic. This setup even works with the devices are on different networks, but the latency becomes more noticeable. We’ve never experienced latency so bad that it made the web keyboard unusable.

If you have the chance to give this feature a try, we’d love to hear your feedback!

Oh, and Meta, if you’re reading this: it would be great if you implemented a system-level version of this so other apps can overcome spatial computing’s keyboard problem too 😉.