A lot of dictation tools takes a little while to get used to as you have to mentally prepare what to say. I’ve thought for a long time that for the population we see a lot who need these this is a skill that is pretty cognitively difficult. Aqua Voice is a new solution to this.

Aqua Voice feels as natural as talking someone at a keyboard. It writes what you say, but it also does what you say.

We achieve these interactions using an ensemble of models fused together in a way that allows them to understand your intent, not just what you literally said.

Examples

Remember that the meeting is on Thursday. - Friday, Friday.

->

Remember that the meeting is on Friday.

(Implicit Correction)

It looks neat. $10/month. See the video demo

There is a lot of evidence for many students (mainly dyslexia in research) that autocorrection software is far more efficient than using a wordpredictor¹. In general, for increasing writing speed, it makes heaps more sense (Note: predictor software can be better if your aim is supporting literacy development) - and I would argue for other areas of AT too ². Using any AI tools (GPT or Bard - but even a lot smaller models will do this), they will “fix” your sentence with a prompt, e.g. “correct this typo sentence including spaces”. But doing this simply, in real-time without user interaction³, is complicated. This fixkey is getting there and worth keeping an eye on ( it needs a shortcut key, which isn’t perfect. It needs to run without that really) https://www.fixkey.ai

---

1. This paper is a bit old now - but was from the University of Cardiff onlinelibrary.wiley.com/doi/10.10… - note it was using Global AutoCorrect which is now part of TextHelp. Good luck on finding it. Others like this are also useful ↩︎

2. I’m convinced autocorrect is needed more in AAC ↩︎

3. Tools like Grammarly are great too - but for sure, you need to use it like a spell checker. I hacked together an OpenAI API backed script to do this and its not as straightforward as I would have liked see it here if you are interested ↩︎