Artificial intelligence has learned to voice photos

(ORDO NEWS) — Early last year, genealogy social network MyHeritage launched Deep Nostalgia, a deepfake web app that brought old photos of deceased relatives to life.

The offer has gone viral, with more than 100 million uses since launch. A new tool called LiveStory offers to add voice as well.

As with Deep Nostalgia, to create a new presentation, users can upload a photo and use machine learning algorithms to turn it into a short animated video. MyHeritage has joined forces with the Israeli D-ID project, and users of the resource have been able to create a talking portrait of a relative that tells a life story.

Of course, the recording does not sound like the real voice of the person in the picture, but you can choose from a library of more than 140 voices with support for 31 languages. In the future, their list will expand.

The story of the person in the picture can either be automatically generated from the family history resource already at the disposal, or compiled by the user himself, who is required to write the text in the first person – it can be divided into chapters and illustrated with additional pictures.

The rest of the work is done automatically: combining images into a video, converting text to speech based on the selected voice model, as well as animating a picture of the relative himself with speech articulation and facial expressions. Depending on the length of the text, all this takes from 30 seconds to several minutes.

The system works equally effectively with black-and-white, colorized or color images, and the resulting video can be further edited if necessary, so that later it can be sent to loved ones.

In the future, it will be possible to add subtitles and change the order of chapters. The tool is available on a freemium model – all features are unlocked by subscription.


Contact us: [email protected]

Our Standards, Terms of Use: Standard Terms And Conditions.