Output from edge-tts is 24k mp3. There were two steps that combined temp files into new mp3s, which was introducing some artifacts and leading to bad final quality. This PR switches to using FLAC format for temporary files, and seems to have improved the audio quite a bit. It's not perfect, but I think the final output is now about equal to the individual mp3's we get back from edge-tts.
Output from edge-tts is 24k mp3. There were two steps that combined temp files into new mp3s, which was introducing some artifacts and leading to bad final quality. This PR switches to using FLAC format for temporary files, and seems to have improved the audio quite a bit. It's not perfect, but I think the final output is now about equal to the individual mp3's we get back from edge-tts.