Adobe Enhance Speech improves the rendering of audio recordings

Adobe Enhance Speech improves the rendering of audio recordings

Artificial intelligence is not only used to create funny or bizarre images from a few words. These technologies are also very useful for modifying and improving all kinds of documents. Pixelmator, for example, uses Apple’s CoreML to enlarge images without much loss of quality.

Adobe has decided to take an interest in sound, with the tool Enhance Speech developed as part of the Shasta project, renamed simply Adobe Podcast. The principle is simple: the service takes a poor audio file, with noise and echo, which it transforms into a file that can be used in a podcast. Small example:

Basic recording with an iPhone carried at arm’s length and with the washing machine running in the background.
The result produced by Adobe Podcast.

Not so bad, even if there is perhaps a somewhat mechanical side to the result (it must be said that the base file isn’t great). However, this tool could be of great help to anyone who wants to get started with podcasting without necessarily having the software (or the know-how, there’s no harm in learning) to clean up recordings.

It’s free to use, but you still need to create an Adobe account (free). Enhance Speech supports MP3 or WAV files up to one hour (or 1 GB). It is possible to download the result. The most hackers can go to the open-source solution Mayavoz available on GitHub.