Audio generation methods and systems
View Patent ↗A method of generating audio assets, comprising the steps of: receiving an input audio asset having a first duration, generating an input image representative of the input audio asset, training a generative model on the input image and implementing the trained generative model to generate an output image representative of an output audio asset having a second duration different to the first duration, and generating the output audio asset based on the output image.
1. A method of generating audio assets, comprising the steps of:
receiving an input audio asset having a first duration,
generating an input image representative of the input audio asset,
training a generative model on the input image and implementing the trained generative model to generate an output image representative of an output audio asset having a second duration different to the first duration, and
generating the output audio asset based on the output image,
wherein the input image and output image each comprise an axis representative of time duration, and the step of generating an output image comprises retargeting the input image along the axis representative of time duration, and
wherein the output image has a larger dimension along the axis representative of time duration than the input image.
2. A method according to claim 1 , wherein the output image has a dimension along the axis representative of time duration which is multiplied by a scale factor.
3. A method according to claim 1 , wherein the input image is an input spectrogram, and the output image is an output spectrogram.
4. The method according to claim 3 , wherein the step of generating an input image comprises performing a Fourier transform on the input audio asset, and the step of generating the output audio asset comprises performing an inverse Fourier transform on the output image.
5. The method according to claim 1 , wherein the generative model is a single-image generative model comprising a generative adversarial network, GAN, having a generator and a patch discriminator.
6. A method according to claim 1 , wherein the step of receiving an input audio asset comprises receiving, from a video game environment, video game information, and the step of generating the output image comprises feeding the video game information into the generative model such that the output image is influenced by the video game information.
7. A method according to claim 1 , further comprising the step of storing the trained generative model on a memory, configured to be accessed to generate further audio assets.
8. A method according to claim 1 , wherein the step of receiving an input audio asset comprises receiving a second input audio asset, and the step of generating an input image comprises generating a multi-channel image, comprising a first channel having an image representative of the input audio asset, and a second channel having an image representative of the second input audio asset.
9. A non-transitory computer-readable medium having stored thereon a computer program comprising computer-implemented instructions that, when run on a computer, cause the computer to implement a method of generating audio assets, comprising the steps of:
receiving an input audio asset having a first duration,
generating an input image representative of the input audio asset,
training a generative model on the input image and implementing the trained generative model to generate an output image representative of an output audio asset having a second duration different to the first duration, and
generating the output audio asset based on the output image,
wherein the input image and output image each comprise an axis representative of time duration, and the step of generating an output image comprises retargeting the input image along the axis representative of time duration, and
wherein the output image has a larger dimension along the axis representative of time duration than the input image.
10. A system for generating audio assets, the system comprising:
an asset input unit configured to receive an input audio asset having a first duration, and to convert the input audio asset into an input image,
an image generation unit configured to implement a generative model to generate one or more output images based on the input image, the output image representing an output audio asset having a second duration different to the first duration, and
an asset output unit configured to generate an output audio asset based on the output image,
wherein the input image and output image each comprise an axis representative of time duration, and the step of generating an output image comprises retargeting the input image along the axis representative of time duration, and
wherein the output image has a larger dimension along the axis representative of time duration than the input image.
11. A system according to claim 10 , further comprising a transform unit configured to perform Fourier transform operations and inverse Fourier transform operations to convert between audio and graphical files, and wherein
the asset input unit is configured to access the transform unit to convert the input audio asset into an input image, and
the asset output unit is configured to access the transform unit to convert each output image into an output audio asset.
12. A system according to claim 10 , further comprising a video game data processing unit, configured to process video game information derived from or relating to a video game environment and feed through to one or more of the asset input unit, the image generation unit and the asset output unit, and the image generation unit is configured to implement the generative model based at least in part on the video game information.
13. A system according to claim 10 , configured to store the generative model on the memory.