OpenVINO is an Intel toolkit for deploying AI models. This particular project is an Intel project using the OpenVINO toolkit to package several existing models as audacity plugins.

> an intel-only solution.

> OpenVino is just DeepFilterNet glued to Whisper.cpp and tied to Intel hardware.

Well, no.

When you want to run a model on a truly wide set of devices, you end up sort of wedged into either ONNX, OpenVINO, TensorFlow Lite, and a few other frameworks.

They're all FOSS, and they're software libraries.

YMMV on which is best, of course, but broadly and widely: where are your users, mostly? Desktop? OpenVINO. Web? TensorFlow. Mobile and desktop? ONNX. This isnt entirely accurate because ex. I reach for ONNX every time because that is what I'm familiar with. All of them make effort to reach every platform, ex. OpenVINO goes supports ARM, and not in a trivial manner.

That all being said, TL;DR:

It is "not even wrong", in the Pauli sense, to imply OpenVINO is Intel-only, and to describe OpenVINO as "just glu[ing a model to inference code]"

You're describing 3 different components (a hardware acceleration library, and inference library, and a model) and suggesting the hardware accelerated inference library just glues together a model-specific inference library and a model. The mastroyshka doll is inverted: whisper.cpp uses openvino to acclerate its model-specific inference code.

The Noise Removal plugin takes a bit getting used to, but I have had great results from it. I don't mean to point to operator error... it's got too many options for someone new to it to tune.

I used to work at Intel doing OpenVINO stuff. Should work on AMD too; it's just not validated for it so there might be quirks.

They have some NVIDIA support in the form of external project: https://github.com/openvinotoolkit/openvino_contrib/tree/mas...

Nvidia has Broadcast which is Windows-only:

https://www.nvidia.com/en-us/geforce/broadcasting/broadcast-...

I ran it on my amd 5950x with an rtx3090. It crashed when I attempted GPU process but ran fine on CPU. YMMV

So use your cpu

It might also work with AMD CPUs too

I don't know if there's an available tool to do it, but "assuming these sources are aligned in time, remove this reference B from that recording A" would be quite a nice undergrad problem. You'd do something like cross-correlate A with B, multiply B by the correlation coefficient and then subtract the result from A. In the frequency domain, because that makes things a little easier.

The next question on the problem would be "Give at least three reasons why this doesn't perfectly remove the reference sound," of course.

I would try openvino, it might get you part of the way there.

Other things to do would be to fix any tape warble or flutter, normalize the volume, do simple things like high pass filtering anything below 75hz (as most voices don't make audible volume at those frequencies, especially children's voices.

Then I would get a spectrum analyzer plugin and see if there are any spots that are clearly music vs children speaking and zap them out.

(Audition is pretty good software for this, you might still be able to find the download and mass unlock serial key that Adobe released for version 3.0 somewhere on the internet, of course, this is only for people who bought it as a perpetual license back in the day and need to activate it now that the activation servers have gone offline, so no pirating it!)

I'm not gonna say it will be perfect but you might do well enough to be able to hear what everyone is saying and it not sound too bizarre.

Try iZotope RX