Implement AI Audio Software using Python
Budget: $250 – $750 USD
The linked scientific paper describes Hifi-Gan 2, an AI based tool able to denoise and de-reverberate an audio file:
https://pixl.cs.princeton.edu/pubs/Su_2021_HSS/index.php
The described tool takes in an audio file and transforms it into an audio that has an increased quality and sounds as it it was recorded in a recording studio.
The requirements:
- Implement software described in the paper using python and publicly available libraries.
- The usage of the software must be as simple as possible. As an input it should take a path to an audio file (wav). The output should be a .wav file that contains the same audio as the input file but cleaned up as described in the paper.
- It must be possible to run it on a local machine without internet access.
- It is expected that you provide everything required to run the software, including a trained model.
- The quality should be on the same level as the ones shown in the samples from the paper (https://daps.cs.princeton.edu/projects/Su2021HiFi2/index.php)
- Delivery should include source code and all data required to run the software (including the trained model)
When evaluating I will:
- Look whether the paper was implemented as described
- Check whether I can reproduce the results
- Check whether the quality is comparable to the one in the samples from the paper
If you are interested please let me know and reply with a summary of the above requirements in your words. I will only consider offers that include such a summary.
https://pixl.cs.princeton.edu/pubs/Su_2021_HSS/index.php
The described tool takes in an audio file and transforms it into an audio that has an increased quality and sounds as it it was recorded in a recording studio.
The requirements:
- Implement software described in the paper using python and publicly available libraries.
- The usage of the software must be as simple as possible. As an input it should take a path to an audio file (wav). The output should be a .wav file that contains the same audio as the input file but cleaned up as described in the paper.
- It must be possible to run it on a local machine without internet access.
- It is expected that you provide everything required to run the software, including a trained model.
- The quality should be on the same level as the ones shown in the samples from the paper (https://daps.cs.princeton.edu/projects/Su2021HiFi2/index.php)
- Delivery should include source code and all data required to run the software (including the trained model)
When evaluating I will:
- Look whether the paper was implemented as described
- Check whether I can reproduce the results
- Check whether the quality is comparable to the one in the samples from the paper
If you are interested please let me know and reply with a summary of the above requirements in your words. I will only consider offers that include such a summary.