The present disclosure describes techniques for differentiable wavetable synthesizer. The techniques comprise extracting features from a dataset of sounds, wherein the features comprise at least timbre embedding; input the features to the first machine learning model, wherein the first machine learning model is configured to extract a set of N×L learnable parameters, N represents a number of wavetables, and L represents a wavetable length; outputting a plurality of wavetables, wherein each of plurality of wavetables comprises a waveform associated with a unique timbre, the plurality of wavetables form a dictionary, and the plurality of wavetables are portable to perform audio-related tasks.

A method of and system for generating audio having pitch attributes of an incoming audio stream. The method comprises receiving a digital audio input. The audio spectrum is analyzed and integrated over segments of digital audio data upon receiving analysis triggers which can be synced with the audio tempo. The integrated spectrum is processed to find peak frequencies in the spectrum and their associated gain stored in a peaks array. The peak frequencies are used to program the oscillators controllable attributes and characteristics. The synthesis is performed upon receiving an analysis clock. A number of digital oscillators are configured with the associated frequency parameters and gain parameters from a peaks array. The oscillators are configured according to the audio pitch analysis and generate an oscillator output at the frequency and gain specified in the peaks array. These oscillator outputs are summed together generating synthesized audio.

An electronic musical instrument includes a playing operator, and a sound source which performs processing of receiving, in response to a user operation on the playing operator, a sound generation instruction according to playing operation information including the pitch information indicating the certain pitch and sound volume information indicating a certain volume, and generating sound according to the certain pitch and the certain volume, based on excitation data generated by multiplying partial data by a window function, the partial data being included in excitation signal waveform data generated based on a plurality of waveform data items which are respectively different from each other in sound intensity in the certain pitch.

A method of and system for generating audio having pitch attributes of an incoming audio stream. The method comprises of receiving a digital audio input. The audio data is analyzed upon receiving an analysis trigger indication which can be synced with the audio tempo. The analysis includes a Fast Fourier Transform of a segment of digital audio data.

The integrated spectrum is process to find a number of peak frequencies in the spectrum and their gain. A number of the peak frequencies are used to program a number of oscillators and the parameters determined during analysis and controllable attributes and characteristics of the oscillators. The synthesis is performed upon receiving an analysis clock. First, a number of digital oscillators are configured with the associated frequency parameters and gain parameters from the peaks array; The number of oscillators configured according to the audio pitch analysis and are generated an oscillator output at the frequency and gain specified in the peak array. These oscillator output are summed together thereby generating synthesized audio.

An electronic musical instrument includes a playing operator, and a sound source which performs processing of receiving, in response to a user operation on the playing operator, a sound generation instruction according to playing operation information including the pitch information indicating the certain pitch and sound volume information indicating a certain volume, and generating sound according to the certain pitch and the certain volume, based on excitation data generated by multiplying partial data by a window function, the partial data being included in excitation signal waveform data generated based on a plurality of waveform data items which are respectively different from each other in sound intensity in the certain pitch.

A musical audio composition is generated based on a content library. The library is a collection of sequences and instruments. Sequences are partial musical compositions, while instruments are groups of audio samples. Instruments are made of audio data and musical data describing the events recorded in the audio. The process begins by reading the library. A new chain is created. A succession of sequences are selected to create a series of segments in the chain. The events in the selected sequences determine the selection of instruments. Algorithms determine the final arrangements and exact modulations of source audio to target outputs. The source audio are modulated, mixed and output as a stream of audio data. Finally the selections and events of the finished segment are output as metadata. An unlimited number of segments can be fabricated in series, each building and evolving from the preceding segments in the chain.

When an instruction is provided for changing a characteristic of a set filter which includes a plurality of partial filters and forms a specified characteristic by combining a plurality of partial filters, a processor performs, as crossfading processing for a first filter and a second filter among the plurality of partial filters, fade-out processing of gradually decreasing a degree of contribution of the first filter to the characteristic and fade-in processing of gradually increasing a degree of contribution of the second filter to the characteristic. As a result, unnaturalness occurring at the time of changing filter characteristics is solved.

A digital-to-analog converter performs a computation process to start the computation based upon the second clock signal with respect to the digital data of music sound if the computation is not under execution when the control signal is outputted by the signal output process, a control process to inhibit the computation based upon the second clock signal from being started with respect to the digital data of music sound until the computation is not under execution when the computation is under execution, and an output process to convert a computation result of the computation process into an analog signal and output the analog signal.

A digital-to-analog converter performs a computation process to start the computation based upon the second clock signal with respect to the digital data of music sound if the computation is not under execution when the control signal is outputted by the signal output process, a control process to inhibit the computation based upon the second clock signal from being started with respect to the digital data of music sound until the computation is not under execution when the computation is under execution, and an output process to convert a computation result of the computation process into an analog signal and output the analog signal.

When an instruction is provided for changing a characteristic of a set filter which includes a plurality of partial filters and forms a specified characteristic by combining a plurality of partial filters, a processor performs, as crossfading processing for a first filter and a second filter among the plurality of partial filters, fade-out processing of gradually decreasing a degree of contribution of the first filter to the characteristic and fade-in processing of gradually increasing a degree of contribution of the second filter to the characteristic. As a result, unnaturalness occurring at the time of changing filter characteristics is solved.