Neural networks for changing characteristics of vocals

A system comprising:

• one or more computer processors;
• and
• one or more computer-readable mediums storing instructions that, when executed by the one or more computer processors, cause the system to perform operations comprising:
• receiving input vocal data, the input vocal data representing input vocals having a first characteristic;
• transforming the input vocal data to an input image, the input image representing frequencies and intensities of the input vocals, wherein the input image has a plurality of pixels with a pixel height and a pixel width, and wherein a first axis is a time, a second axis is a frequency, and an amplitude is represented by an intensity of a pixel of the plurality of pixels;
• processing the input image using a first convolutional neural network (CNN) to generate a first output image;
• processing the first output image, using a second CNN, to generate a second output image;
• determining a loss between the input image and the second output image;
• and
• training the first CNN based on the determined loss.

A non-transitory computer-readable storage medium including instructions that, when processed by a computer, configure the computer to perform operations comprising:

• receiving input vocal data, the input vocal data representing input vocals having a first characteristic;
• transforming the input vocal data to an input image, the input image representing frequencies and intensities of the input vocals, wherein the input image has a plurality of pixels with a pixel height and a pixel width, and wherein a first axis is a time, a second axis is a frequency, and an amplitude is represented by an intensity of a pixel of the plurality of pixels;
• processing the input image using a first convolutional neural network (CNN) to generate a first output image;
• processing the first output image, using a second CNN, to generate a second output image;
• determining a loss between the input image and the second output image;
• and
• training the first CNN based on the determined loss.