AudioSR: Versatile Audio Super-resolution at Scale

Pass your audio in, AudioSR will make it high fidelity!

Work on all types of audio (e.g., music, speech, dog, raining, ...) & all sampling rates.

Share your thoughts/samples/issues in our discord channel: https://discord.gg/HWeBsJryaf

Change Log

2023-09-24: Add replicate demo (@nateraw); Fix error on windows, librosa warning etc (@ORI-Muchim).
2023-09-16: Fix DC shift issue. Fix duration padding bug. Update default DDIM steps to 50.

Commandline Usage

Installation

# Optional
conda create -n audiosr python=3.9; conda activate audiosr
# Install AudioLDM
pip3 install audiosr==0.0.7

Usage

Process a list of files. The result will be saved at ./output by default.

audiosr -il batch.lst

Process a single audio file.

audiosr -i example/music.wav

Full usage instruction

> audiosr -h

> usage: audiosr [-h] -i INPUT_AUDIO_FILE [-il INPUT_FILE_LIST] [-s SAVE_PATH] [--model_name {basic,speech}] [-d DEVICE] [--ddim_steps DDIM_STEPS] [-gs GUIDANCE_SCALE] [--seed SEED]

optional arguments:
  -h, --help            show this help message and exit
  -i INPUT_AUDIO_FILE, --input_audio_file INPUT_AUDIO_FILE
                        Input audio file for audio super resolution
  -il INPUT_FILE_LIST, --input_file_list INPUT_FILE_LIST
                        A file that contains all audio files that need to perform audio super resolution
  -s SAVE_PATH, --save_path SAVE_PATH
                        The path to save model output
  --model_name {basic,speech}
                        The checkpoint you gonna use
  -d DEVICE, --device DEVICE
                        The device for computation. If not specified, the script will automatically choose the device based on your environment.
  --ddim_steps DDIM_STEPS
                        The sampling step for DDIM
  -gs GUIDANCE_SCALE, --guidance_scale GUIDANCE_SCALE
                        Guidance scale (Large => better quality and relavancy to text; Small => better diversity)
  --seed SEED           Change this value (any integer number) will lead to a different generation result.
  --suffix SUFFIX       Suffix for the output file

Troubleshooting

RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 64 but got size 63 for tensor number 1 in the list. This error is caused by the duration padding. The oringinal pad_duration calculation is not accurate. Just round up the duration to the nearest multiple of 5.12 in the utils.py file and everything will be fine.
```
  # PATH_TO_ENV/.conda/envs/audiosr/bin/python/site-packages/audiosr/utils.py

  # line 195
  if(duration % 5.12 != 0):
      pad_duration = int((duration // 5.12) + 1) * 5.12
  else:
      pad_duration = duration
```

TODO

Add gradio demo.
Optimize the inference speed.

Cite our work

If you find this repo useful, please consider citing:

@article{liu2023audiosr,
  title={{AudioSR}: Versatile Audio Super-resolution at Scale},
  author={Liu, Haohe and Chen, Ke and Tian, Qiao and Wang, Wenwu and Plumbley, Mark D},
  journal={arXiv preprint arXiv:2309.07314},
  year={2023}
}

ghnmqdtg/AudioSR