Whispp lets people with speech impairments speak in real time using AI

The Dutch start-up Whispp uses AI to support people with speech impairments. For example, its application converts soft voices into speech that is easier to understand while retaining emotion and intonation. At the MWC in Barcelona, Whispp is presenting a new function that also offers AI support in personal conversations.

Similar to simultaneous translation apps, you speak directly into your cell phone. The other person then needs headphones or speakers connected to the cell phone to hear the boosted voice. In the future, a special loudspeaker could be developed that has the lowest possible latency to the Whispp app, says CEO Joris Castermans in an interview with heise online at the MWC. At the moment, you still have to rely on conventional playback devices.

The lowest possible latency is particularly important for personal calls, says Castermans. That's why the new function will also switch to local AI processing of voice input from the summer. So far, this is still happening in the cloud. In a demo at the MWC, there was a processing latency of 220 milliseconds with a wired speaker. Although this is a little longer than we are used to from phone calls via Bluetooth headphones, it is sufficient for a natural conversation.

Audio directly to audio

The fact that it works so quickly is due to the way Whispp works. Unlike other applications, the AI converts audio directly into audio. The program therefore saves itself the intermediate step of first converting speech into text and then back into speech.

If you wish, you can still train Whispp with your own voice. This is useful for people who have lost their voice as a result of an illness such as throat cancer, but still have voice recordings of their old voice. If you enter them into Whispp, the app creates a "voice clone" – which is supposed to sound particularly authentic.

According to the Dutch developer, Whispp can be used regardless of language as long as the user does not speak a particularly strong dialect and articulates well. The app is primarily aimed at people with speech impediments and people who stammer – because it is often easier for them to speak fluently in a whisper.

Also for (sick) business customers

The Whispp application is already available for Android and iOS, and a desktop version for Windows devices is in the works. According to Castermans, Whispp has currently been downloaded 10,000 times. This means that 1,500 people use the app every month.

In order to expand its audience, Whispp now also wants to attract business customers. For example, the company's AI technology could enable confidential business conversations to be held in hushed tones, even in public places. In addition, employees in service centers could use Whispp to make phone calls even if they are hoarse after illness.

The mobile app can be downloaded and tried out free of charge. Voice messages are free, but those who want to make calls have to pay 20 euros per month. A few more euros per month could be added for simultaneous calls, says CEO Castermans. In the long term, however, the aim is to license the technology to other manufacturers anyway.

(dahe)