Audio technology has come a long way, and iZotope continues to push the boundaries with their cutting-edge tools. One of their latest innovations is the Music Rebalance module in RX 7. This powerful tool allows users to boost, attenuate, or even isolate individual musical elements from audio recordings, without the need for access to the original stems. In this article, we will delve into the fascinating technology that makes RX 7 Music Rebalance possible and explore its various applications.
The Evolution of Intelligent Audio Technology
iZotope is committed to developing intelligent audio technology that empowers creativity and makes the impossible possible. With each new release, they introduce groundbreaking features that redefine what can be achieved in audio production. Tonal Balance Control in Ozone 8 and Track Assistant in Neutron are just a few examples of their assistive audio tools. Music Rebalance in RX 7 is a natural progression of their neural network-based source separation technology.
What is Music Rebalance and How is it Useful?
Music Rebalance is a game-changing tool that allows users to adjust the levels of individual musical components in a mix, even without access to the original stems. This functionality opens up a world of possibilities across various applications. Mastering engineers can now re-level sub-par mixes without the hassle of coordinating file transfers with mixing engineers. Producers can sample music freely, adjusting percussions to avoid interference with drums or isolating vocal phrases for remixes. Music students can accentuate specific musical elements to aid in transcriptions, and even karaoke enthusiasts can remove vocals from their favorite songs to generate instrumentals that closely resemble the original tracks.
How Does Music Rebalance Work?
Music Rebalance harnesses the power of machine learning technology to separate different musical sources in a mix. Under the hood, the audio is processed through several neural networks, each trained to identify and isolate a specific musical source of interest, such as vocals, bass, or percussion. These neural networks provide information on the presence of each source at different times and frequencies in the audio. Based on this information, Music Rebalance allows users to control the gain of each musical source, along with a remainder source called “other,” relative to their original amounts in the mix.
The performance of the source separation technology heavily relies on the quality of the training data. To ensure accurate and reliable results, iZotope meticulously trained each source isolation network on isolated stems for each source type. For example, the vocal network was trained on a variety of singers with different timbres and vocal ranges. By training on diverse genres and mixes, the networks were able to generalize and separate different voices in a mix effectively.
The Training Process
During the training phase, each custom mix consisting of vocals and instrumental tracks was used as input, while the soloed vocal stem was considered the ideal output. This approach allowed the network to learn to output a mask that corresponds to the vocal components in any given mix. To cover a wide range of use cases, iZotope trained on vocals from various genres and mixed them with instrumental stems from different genres. This diverse training set ensures that the networks are capable of separating voices effectively across different musical styles.
To further enhance the training process, iZotope employed augmentation techniques. These techniques involved adjusting the amount of each source in the training mix, allowing the networks to learn separation even with different source levels. By avoiding overfitting to specific types of mixes, the networks can deliver reliable results across a broad range of audio recordings.
The Future of Machine Learning in Audio Processing
Machine learning and neural networks have found a prominent place in iZotope’s audio processing tools, and they are excited to explore their potential in the future. With Music Rebalance as a stepping stone, iZotope envisions a future where neural networks are leveraged in audio restoration, synthesis, and auto-mixing. The ability to separate individual instruments using machine learning is just the beginning, and iZotope looks forward to the new possibilities that will arise from further advancements in this field.
Conclusion
iZotope’s RX 7 Music Rebalance module is a groundbreaking tool that empowers users to manipulate individual musical components in a mix with ease. Through the power of machine learning and neural networks, Music Rebalance provides unprecedented control over audio recordings, even without access to the original stems. Whether it’s re-leveling sub-par mixes, sampling music, aiding in transcriptions, or creating karaoke instrumentals, Music Rebalance opens up a world of possibilities for audio professionals and enthusiasts alike. With the rapid advancements in machine learning, the future of audio processing looks promising, and iZotope continues to lead the way with their innovative technologies.