Sound super-resolution is a technology that allows you to “finish” high-frequency sound components. AudioSR’s diffusion generative model is capable of delivering high-quality super-resolution for a variety of audio types, including sound effects, music and speech. AudioSR is capable of upsampling any input audio signal in the range from 2 kHz to 16 kHz to high-resolution audio with 24 kHz bandwidth (48 kHz sampling rate). Extensive objective evaluation of various tests demonstrates the high results achieved by the proposed model.
- Unpack the archive. The path (folder names) must not contain spaces or Cyrillic characters.
- Run the file “run.cmd”.
- Select a model (voice – number 1, music – number 2)
- Enter the path to the file. (Important! The path must not contain quotes!)
- The result will be located in the output folder.