I stopped using Descript 6 months ago. I stopped using the product because there was one element that made it incredibly frustrating to use:

The fact that you cannot use existing audio to train a voice. This makes it incredibly difficult to regenerate audio for projects that have already been recorded, perhaps on different audio equipment and in different environments than what the editor has access to.

The current system where training your voice is also folded into the verification process is irritating, because it means that I can't submit that perfect voice sample that matches the existing audio, or indeed, just train the voice sample on the existing audio that I want transcribed. I to submit audio that doesn't match and as a result the regenerated audio is woefully inaccurate.

I think this tool has a lot of potential and I hope this feature can be added to make it more versatile and flexible.