‘DeRoom’: Accentize’s Intelligent Real-Time Reverb Reducer based on Machine Learning Techniques
’DeRoom’ removes unnecessary reverbs and room resonances by means of its internal machine learning algorithm in real-time.
Recent advances in the area of machine learning have evidently shaped the way we handle, understand, and process data. New theoretical insights, the rise of freely available programming libraries, and the increasing access to computational resources provide a huge set of new possibilities to tackle data processing problems from a different angle. Accentize focuses on applying cutting-edge machine learning methods to the area of audio signal processing and closes the gap between state-of-the-art research and productive implementations. The newly released Accentize’s ‘DeRoom’ plug-in will help you to reduce unnecessary reverbs and room resonances from audio recordings in an easy and automated fashion.
‘DeRoom’ is a real-time reverb reducer working under specially trained artificial neural network that analyze room acoustics. The artificial neural network applied to ‘DeRoom’ has been trained on many different room scenarios in order to be able to separate direct sound from reflection components. With the help of machine learning techniques such as the artificial neural network, ‘DeRoom’ can detect and remove reverbs and room resonances from any kind of room environments in real-time.
The DeRoom Engine
‘DeRoom’ is designed to detect and analyze reverbs and room resonances in three different simulated room types, which are Small, Medium, and Large. The algorithms for reverb detection have been developed in individual optimization sessions and can be easily compared against each other among these three types of room. Therefore, these three algorithms have been optimized for different scenarios of room-size. As a result, users can make fine-tune their results by selecting between the three different room types.
Controls Over Reverbs
Don’t forget the ‘DeRoom’ is running under machine learning algorithm, so users do not really need to estimate and set room-timing constants by hand. The algorithm of ‘DeRoom’ will figure it out internally in an automatic fashion, so users only need to set the reverb reduction amount and let the neural network do the rest.
The ‘Reduction’ parameter defines how much room-sound reduction you are aiming for. 100% Reduction will result in complete removal of the room sound and leave only the dry direct signal. 0% Reduction means no room-sound reduction at all.
The ‘Make-up’ adds a broadband gain in order to compensate for possible level loss of the target signal after processing.
The Sound from 'DeRoom'
The drum sound from ‘DeRoom’ varies depending on the room-size settings and I realized the bigger the size the smaller the reverb tails. Each algorithm of ‘DeRoom’ automatically detects reverbs and room resonances based on the simulated room scenarios and this results in the different ratio between direct sound and reflections in each scenario. The filter cut the room sound in accurate and precise manner without touching other parts of the signals, so I could get clear direct sounds from the drum track without any modulation or coloration over the original signals. Again, the separation of direct sound from reflections is quite easy with 'DeRoom' plugin and it guarantees ultra reality in the processed sound. In addition, the ‘Sensitivity’ control can be used to change tonal quality of the target signal so users are able to shape the reflective walls as desired using this control.
The artificial neural network applied to ‘DeRoom’ has been trained on many different room scenarios in order to be able to separate direct sound from reflection components, so ‘DeRoom’ plugin will handle almost all kind room sounds (reverbs and room resonances) from various room environments.
With ‘DeRoom’ You can do:
- Reduce reverb and room resonances from audio recordings in real-time
- Separate direct sound from reflection components
- Shape room acoustics to your taste
- Tackle different kinds of reverbs without introducing many artifacts
- and more.
€41 EUR (Release discount until 27 March 2020)
*Note that a 7-day, fully-functional trial version for macOS and Windows can be downloaded free: here
Accentize Recent advances in the area of machine learning have evidently shaped the way we handle, understand and process data. New theoretical insights, the rise of freely available programming libraries and the increasing access to computational ressources provide a huge set of new possibilities to tackle data processing problems from a different angle. We focus on applying cutting-edge machine learning methods to the area of audio signal processing and close the gap between state-of-the-art research and productive implementations.
View this article in Japanese version: 1
< Copyright ⓒ The Music Telegraph :: Prohibit reprinting and redistribution >
Product News (popular articles)