Zerebrix, LLC (March 10, 2021) – Today, Zerebrix, the experts in advanced audio technology, released AudioMint, a new product that can either isolate vocals or remove vocals. AudioMint sets itself apart from other vocal isolators in that it is able to generate audio sources from the ground up to fill in missing gaps in isolated sources. This results in isolated vocals which sound more natural than those obtained via other vocal isolators.
The task of isolating audio sources is a heavily researched topic in both academia and the audio industry. The leading techniques for accomplishing this task use machine learning to train a neural network to learn how to split audio up into its component parts. This is commonly done by training the network to minimize the difference between an isolated audio sample and that same sample mixed with other instruments. Most methods are only able to rely on the information that is presented to the neural network for inference on what the isolated sample will sound like. This method of minimizing differences is the same one used by the open-source software Spleeter and many other websites.
The issue with these methods is that sometimes sounds from sources in overlapping frequencies cancel each other out, and therefore some information is lost. This makes it impossible for these older methods to completely and convincingly isolate audio sources.
Original audio sources are damaged during mixing due to a phenomenon called phase interference. When two waves in overlapping frequencies interact it causes the two waves to couple, and therefore causes a loss of information in the signal. AudioMint identifies this information loss by using context clues in the area around the phase interference and generating its own version of what it “thinks” the original source sounded like. The resulting isolated source will not be the same as the original in this case, but in most cases, it will sound close enough to the point where it is indistinguishable from the original source.
We believe AudioMint produces isolated vocals that are unrivaled in quality. However, you do not need to take our word for it, we have provided the below samples so you could compare AudioMint to the open-source vocal isolator, Spleeter.
Sample 1 (Every Time The Son Leaves by Telepathic Teddy Bear)
As you can see from sample 1, both AudioMint and Spleeter were able to isolate the vocals, but the Spleeter version sounds distorted. This is due to the loss in phase information, causing a whooshing effect in the Spleeter version. However, the AudioMint version produces a more accurate result that sounds much closer to the original source than the Spleeter version.
Sample 2 (My Little Kingdom by Golden Duck Orchestra)
In sample 2, the guitar from the background makes the Spleeter version sound even more distorted resulting in vocals which are completely unusable for most projects. The lack of phase information results in a hissing noise reminiscent of the sound you hear in a webcam call. However, in the AudioMint version, we again get a result that is much more faithful to the original and cleaner of artifacts.
Want to try AudioMint for yourself? You can try it out for free here.
Zerebrix produces audio and video editing software using handcrafted state-of-the-art AI algorithms. Our goal is to enhance the lives of music producers, video editors, and audio engineers by giving them the tools necessary to produce professional quality work.
Founded in 2021, Zerebrix is based in Los Angeles, California. To learn more, visit us at www.zerebrix.com.
For more information, please contact:
Sam Baron, Chief Operating Officer, firstname.lastname@example.org
Zerebrix, LLC (March 10, 2021) – Today, Zerebrix, the experts in advanced audio technology, released AudioMint, a new product that can either isolate vocals or remove vocals. AudioMint sets itself apart from other vocal isolators in that it is able to generate audio sources from the ground up to fill in missing gaps in isolated…