Inspired by the Overcast variant of this kind of tool, I built a very basic browser based version using canvas and various web audio/video apis. It was fun to build over a couple weekends, but it ended up not being usable broadly due to speed (runs in realtime linear) and browser limitations of file export types (webm in chrome). If ffmpeg could reliably run in wasm, there could be alternative approaches. I concluded after I built it, I should make a headless non-browser version and it would be more usable, but haven't gotten around to it.