Мы тестируем новую версию сайта. Эта версия обновляться не будет. Приносим извинения за временные неудобства.
We are testing a new version of the site. This version will not be updated. We apologise for any temporary inconvenience.
Estamos probando una nueva versión del sitio. Esta versión no se actualizará. Rogamos disculpen las molestias.

Site statistics

35225 facts from 177 countries related to 1198 phenomena have been registered in Archive. 2834 of them were solved, another 11033 are under verification for compliance with one of the 322 versions.

0 facts have been added for last 24 hours.

Share your story

You are in "News" section

This section accumulates world news published in the media and concerning facts, phenomena and/or versions.

Появился алгоритм, который меняет речь людей на видео

Added	Mon, 23/09/2019
Источники	ohadf.com knife.media
Дата публикации	Fri, 07/06/2019
Версии	Mounting Deliberate falsification

Researchers from Stanford and Princeton universities have created an algorithm that distorts a person's speech on video. For the credibility of the program not only adjusts voice and changes facial expressions, and the length of the roller.

According to the developers, their program is arranged this way: from the video it allocates the audio track, splitting human speech into phonemes. Further ON creates a 3D model of a human face, scanning as he moves his lips, saying the words. Then the machine finds wisely — sounds that look alike when lip movement and uses them to create facial expressions for the new phrase. In the end of the two spoken words, the algorithm generates a third.

Translated by «Yandex.Translator»