September 7, 2022
Neutone for Mac is going off-beta today, fixing major limitations and bugs. We are happy to see a number of artists have started using the plug-in such as BIGYUKI who is working on his upcoming work using Neutone. Since our beta release in May, we’ve expanded the model library with more RAVE and DDSP models including a beautiful-sounding Choir model and authentic sounding Shakuhachi model.
Neutone is no longer limited to a 48kHz sampling rate or 2048 buffer size. We’ve done extensive testing on major Mac OS on both M1/Intel architectures. Unfortunately, the Windows version needs a little more time to develop and we will let you know as soon as it is ready for testing!
Neutone is a plug-in for DAWs compatible with VST3 and Audio Unit (AU), which means we support most DAWs around. Here we’ll show you how to use it in one of our favorite DAWs, Ableton Live.
After installing Neutone with the installer, the VST and AU versions will be installed in the system plugin folders. If the DAW doesn’t find them, please make sure in the preferences that these default locations are scanned by setting the following options On.
Once that’s done, you should see Neutone under Plug-Ins > VST3/Audio Units> Qosmo
Neutone can work on a recorded track, MIDI instrument or live audio input. To use in an audio track, drag and drop the plug-in to an audio track and you’ll see Neutone loaded to the Audio Effects list in the bottom. Now you are ready to use Neutone by clicking the spanner icon.
You will first see the list of available AI models.
Pick any one of the models you are interested. Click download and use to enable the model and you’ll see the main plug-in window:
If you have an experience using other plug-ins, you’d be familiar with the big knobs down below:
There are also some “custom settings” unique to each model. In the case of timber transfer models (RAVE and DDSP as of now), these typically control the way the model respond to the input by changing the behavior of the latent space. It’s difficult to explain what they do systematically but try tweaking them until you get a good output.
Most of the models currently available for Neutone are timbre transfer models, which transforms an input sound timbre to another. Typically you’ll get better result with single instrument input (or voice) and with DDSP models, you should use monophonic (no chords) input.
We also saw better results when the input sound is filtered towards certain frequency range where the model is more sensitive to responding. It is also a good idea to use a limiter or compressor to saturate the dynamic range for better responses. Try for example using set of effects like the following:
We hope we outlined the basics for the uses interested in experimenting with the Neutone plugin. While we are hoping Neutone would become a great tool to expand your creative potential in music production with the array of existing models, we’d love for you to think about building your own models using our notebooks as a starting point. See our previous post regarding this potential. With the current RAVE and DDSP technologies, you will need to record ~2 hours of training data (audio recording for the target instrument in a studio environment) but the training process is fully automated with the notebooks we provide.
Once your model has been wrapped in Neutone SDK, you can read it locally using the “load your own” button in the model list page. Select the .nm file, which contains the model.
We look forward to hearing about your experiences building models and using it in your performances!
We would love to host your models and share it with the musicians and creators who loves using Neutone. Please get in touch with us through this link. We’d be thrilled if you could become a Neutone model developer! Please also note that there is an active Discord community for Neutone which you can join through this link.