Sentient is a new AUv3 Synthesizer for iOS and macOS, powered by a neural network that listens to everyday sounds and uses them to build a personalized wavetable.
Many new synthesizers for iOS have hit the market in recent weeks and months. Many offer familiar features, albeit in new packaging. However, Valery Mironov, developer of the Cellular synth, shows that you can do other things with his latest release.
Sentient is his second release and a Synthesizer powered by a neural network.
Sentient Synthesizer
Some likely know the developer for his distinctive Cellular Synthesizer AUv3 plugin for iOS and macOS, based on cellular automata
Sentient is another wavetable Synthesizer plugin, but it takes a different approach to generating the table and sound. It’s powered by a neural network that listens to everyday sounds and uses them to build a personalized wavetable for each user.
Like Cellular, the new app features a highly distinctive user interface—more playful and less classic or generic than what is commonly seen today.
The top section has a controller with many eyes that represents the neural network learning system. When the eyes are open, the system learns; when they are closed, learning is disabled. You can easily turn it off with a long tap.
Dragging vertically changes the learning rate, while dragging horizontally controls how strongly the winning neuron influences its neighbors. The number of open eyes and the pupils’ width describe the influence. Clever design. While learning, the neural network visibly `swells`.
When a new sound is analyzed, the neuron that best matches it updates its internal representation and `swells` more than the others. Nearby neurons also update, but less strongly, forming a smooth sound space inside the network.
Neural Network Synth
Then, two grids represent the two hemispheres of the neural network, and each cell corresponds to a neuron. Tapping a cell selects the neuron whose waveform it will use. You get an instant representation with the orange waveform that shows the fragment stored by that neuron.
The controller below in Sentient with the planets is basically the neural network’s memory. It lets you move through time and browse stored states. It includes save, delete, and close functions.
The planet interface represents time symbolically. The crescent shadow shows the time of day, the moon’s position indicates the day of the month, and the satellite’s position represents the orbital period of an artificial body similar to the ISS.
Below the neural grids are six synthesis settings: oscillator waveform selection with two sliders for the frequency and volume, and the others for modulation (depth, rate). You can route the modulation to pan, drive, filter, and resonance.
There is also a star slider that controls the filter with filter selection (horizontally) and drive adjustments (vertically). The area to the right controls filter frequency and resonance.
With the eight figures at the end of the app, one might think it involves modulation. But it doesn’t. This is a keyboard based on eight keys, represented by Lissajous figures, each corresponding to a note. These are curves created by combining two harmonic oscillators.
First Impression
Sentient looks really lovely and eye-catching. That’s also why I noticed the app on the developer’s IG; judging by the post, it seems to have caught a lot of attention.
I can’t say how the synth sounds, as there isn’t a demo available yet. However, the concept is unique and sparks curiosity.
Sentient Synthesizer is available now for $5,99 and runs as a standalone and AUv3 plugin on iOS (iPhone and iPad) and macOS (native Apple Silicon).
More information here: Valery Mironov / App Store



Be the first to comment