Exploring assistive text to speech in 2026

In 2026, assistive text to speech has largely solved the problem of sounding human: modern AI voices are expressive, natural and available in dozens of languages. The challenge that remains is speed. Spoken conversation moves far faster than anyone can type, so the next step for assistive TTS is helping people keep up in real conversations, not just speak a sentence once it’s written. This article looks at where the technology stands and what we’re exploring in a new conversation-focused app now in early development.

Where assistive text to speech is today

A few years ago, the voice was the weak link. Synthetic speech sounded flat and robotic, and many people who relied on it felt it didn’t represent them. That has changed. Today’s neural voices carry warmth, rhythm and emotion, and streaming technology means speech can start playing almost as soon as you tap speak.

For many people, that’s enough. If you know what you want to say, a fast, natural voice and a library of saved phrases cover most of daily life. That’s what MyVoice is built for, and it will stay that way.

But conversation isn’t a series of prepared statements. Someone asks a question, waits, and then the moment passes. By the time a reply is typed, the conversation has often moved on. People who use AAC describe this again and again: being talked over, being asked to “just nod”, or having someone else answer for them. The gap isn’t the voice any more. It’s the pace.

The conversation gap

Typical spoken conversation runs at well over a hundred words a minute. Typing on a phone, especially one-handed, with tremor, or on a day when fatigue is high, is a fraction of that. Better voices don’t close that gap. What closes it is reducing how much a person has to type before they can say what they mean.

That’s the question at the heart of our new project: what would a text-to-speech app look like if it was designed around conversation rather than around typing?

What we’re exploring

We’re prototyping a separate, conversation-focused app alongside MyVoice. It’s at an early stage, but a few ideas are shaping it.

Listening as well as speaking

Most AAC apps only speak. Our prototype can also listen. With the user’s permission, it transcribes what the other person says and shows it on screen as part of the conversation. That alone helps: you can see the question you’re answering, and you don’t lose track if you were concentrating on typing.

Because listening involves other people’s voices, privacy comes first. Audio only leaves the phone through a speech service the user has knowingly chosen, and transcripts and conversation history stay on the device.

Reply suggestions that stay put

Once the app knows what was said, it can suggest replies. The prototype offers a small number of reply buttons, drawn first from simple templates, then from the user’s own past replies, and only then from AI. Suggestions start forming while the other person is still talking, so they’re often ready by the time they finish.

One design rule matters more than it might seem: suggestion buttons never move under your finger. For someone with a tremor or limited motor control, a button that jumps just as you reach for it is worse than no suggestion at all. Each suggestion keeps its slot.

Turning short typing into full sentences

Many people who type to speak already use shorthand. The prototype builds on that. It expands abbreviations (“cn u” becomes “can you”), predicts words offline, gently corrects spelling, and can turn short, clipped input into a full sentence.

The rule here is just as important: expansions stay close to what was typed. The app helps you say what you meant, faster. It doesn’t decide what you meant.

One tap to speak, always

Some things should never be slow. In the prototype:

• Speaking is one tap, with no confirmation step in the way.
• Quick replies like Yes, No, Wait and Repeat? are always one tap away.
• Tone and volume can be adjusted, from a polite request to an urgent one, and from a whisper to a loud voice in a busy room.
• An urgent screen keeps large, one-tap phrases ready for moments when there’s no time to type at all.
• Speaking never waits on the network. Natural AI voices stream as they’re generated, but if the connection is slow or missing, the phone’s own voice takes over instantly. Lines you’ve said before are remembered, so repeating them is immediate, even offline.

Designing for real bodies and real days

Assistive technology has to work on bad days as well as good ones. The principles guiding the prototype reflect that:

• Thumb-first layout. The controls you use most sit in the bottom third of the screen, within easy reach when holding the phone in one hand.
• Large, clear targets. Generous touch targets, high contrast and a typeface designed for legibility, with full support for screen readers and the system text size.
• Nothing is lost by accident. Vocabulary is hidden rather than deleted, and edits need a deliberate action, so a stray tap can’t undo someone’s carefully built phrases.
• Lower-effort options. A planned phrase board will make common phrases available for days when typing takes too much energy.
• Everything works without AI. Suggestions and rewrites are a bonus, never a dependency. The core of the app, typing and speaking, works fully offline.

What AI should and shouldn’t do

AI is powerful here, and it’s also easy to get wrong. A reply suggestion that sounds fluent but isn’t what someone meant is not help. It puts words in their mouth.

So our approach is that the person is always the author. AI offers options; it never speaks on anyone’s behalf without a tap. Suggestions learn from the user’s own replies, so over time they sound more like that person and less like a generic assistant. And the simplest tools, like quick replies and saved phrases, always come first.

What’s next

The conversation-focused app is still an early prototype on Android, and there’s a lot to learn before it’s ready for anyone to rely on. The most important next step is testing with the people it’s for: AAC users, their families and the speech and language therapists who support them. Real conversations, real environments and real phones will shape what it becomes.

MyVoice isn’t going anywhere. It stays the simple way to speak, with the same natural voices and your saved phrases. The new app is being built as a companion for keeping up in live conversation, and it will use the same voices so that you can keep the voice that feels like yours.

If you use AAC, support someone who does, or work as a speech and language therapist and want to share what would make the biggest difference in conversation, we’d love to hear from you through our support page.

← Back to blog