With Project Sound Lift, Adobe has introduced another AI-supported technology that is able to separate voice recordings of a video into different tracks (i.e. voices, non-speech noises and other background noises). This means that an old but fundamental problem may no longer be a problem in the future. Namely, that audio recordings can be split into individual sources with their own audio tracks again "without leaving any residue". Such methods have already been increasingly researched in AI projects in recent years, but the practical, commercial application was, as usual, not necessarily at the forefront of research, and it is usually not everyone&s cup of tea to compile and install the corresponding tools themselves and then use them as a command line program.
Adobe&s Project Sound Lift, on the other hand, is intended to be a one-click solution that uses AI to independently optimize, transform and ultimately control speech and sound.
Incidentally, part of Project Sound Lift can already be found in Premiere Pro, where it can already reduce background noise in the function "Improve speech".
The even more advanced Project Sound Lift has now been presented to MAX in Japan as part of the Adobe Sneaks. (In such Sneaks sessions, Adobe engineers and researchers provide insights into prototype ideas and technologies that have the potential to become important components of Adobe products in the future).
Project Sound Lift will soon recognize a wide range of audio events from everyday life, making it possible to split speech, applause, laughter, alarms, crowds and various other ambient sounds into different tracks. Each track can then be individually controlled in a virtual mix. This means that the sound of a clip can be individually controlled even after shooting, as if it were a multi-channel recording.
Adobe has not yet announced when Project Sound Lift will be available for end users. As a rule, however, the functions teased in such sneaks are followed by corresponding products after just a few months...