Windows – How to extract clear voice from a noisy audio file

With Audacity, you will be able to achieve close to everything you want to achieve provided you understand how to work audio. It is important you need to be able to visualize sound in terms of identifying the different frequencies. Frequencies are like the audio equivalent of an image's histogram, and to achieve "HDR" audio, you will need to blend in multiple tracks that has emphasis on different frequencies.

What I recommend for you to do is this (assuming you got two sources of audio):

1. Line up all the individual sources of audio - they should sync or be as close to syncing as possible from start to end.
2. For each source of audio, in Audacity, split it up into four tracks each. You should now have eight tracks, 4 from each original audio stream.
3. Clean up noise for only ONE track of each audio stream, and use the Equalizer function to best boost the important audio in this track (e.g the speaker).
4. Use the Equalizer function, adjust the other 3 so the emphasis is on the lower band freq, middle band freq, and upper band freq.
5. To recap, you should have 8 tracks in a single project in Audacity, 4 from each audio source, 1 of the 4 is denoised and important audio boosted, and the other 3 cover a range of lower freq, middle freq, and upper freq audio.
6. On the left hand side of each track that's showing the waveform, you should see a volume adjuster - this comes the most time consuming part. Adjust the volume for each track individually as you play the the entire project over and over again. Fine-tune it to your desired results.
7. Export as the audio format of your choice - you should now have a piece of HDR audio.

Again, imagine how a photo does HDR - by combining 3 or more photos (of the same shot) with shadow details, with highlight details, and also boosting mid-range tonal details. Similarly, with audio, you combine 3 or more audio tracks that covers the lower freq, middle freq, and upper freq, and have another track that covers critical audio.

Good luck - it will take some work, but I am confident with this method it will get you if not what you want, at least very close to it.

I ended up using Voicemeeter. It has a pretty easily understandable user interface. Didn't work at first, but I managed to pick the right settings and after a few system restarts it's all working just about as I drew on my diagram in the question. The app needs to be running for audio routing to work and you can close it to return to normal without having to restart if you configure it correctly. Did I mention there are no sound artifacts or delays introduced by this solution? :)