Another cool to experiment would be to add an Audio refiner as an optional Step 5 – with the sole goal to make the generation sound as good as possible. Resemble Enhance would fit in well there:

→ View original post on X — @reach_vb