A couple of DeepLearnig experts have come up with an -in our eyes- extremely good idea and made it available for free via GitHub. The "Deep Learning Tools for Audacity" is a software framework that allows deep learning researchers to easily integrate their own PyTorch models into the open source DAW Audacity. This allows AI audio researchers' tools and models to be tried and used by anyone.
Audacity is a free and open source digital audio workstation (DAW) that can claim over 100 million downloads since 2015. As an audio editor, Audacity is also quite popular among video editors, especially since the program is available for all relevant operating systems.
Using the "Deep Learning Tools for Audacity" software framework, developers can upload their trained PyTorch model to HuggingFace's Model Hub, one of the largest AI model databases on the Internet. Afterwards, the model is accessible through Audacity's user interface and can be loaded similarly to traditional plug-ins:
In a separate 
arxiv paper/PDF, the developers show some application examples for both end users and neural network developers.
To try it out for yourself, you (still) have to download a special version of Audacity, since the framework has not yet been integrated into the main version of Audacity. According to the developers, however, the feature should end up in the main version soon - if you want to follow the progress, you can check out the corresponding pull request on Github. The download of the special alpha version can be found on the linked developer page for PC and Mac (but incomprehensibly not for Linux).
In our eyes, this is an extremely exciting development that we would also urgently wish for video editing software. Therefore, dear software manufacturers, now please also something comparable for video software, a ONNX importer would be enough for us for now...
;)