found this workflow and had to share - Before you start: - Don't generate during peak times. Quality is too bad to even fix. (credits) Stem Splitting: - Option 1: Mvsep - Music & voice separation (ree) - Option 2: Kits AI - Vocal Remover (Paid) - Bonus tool: Adobe's 'Enhance Speech v2' (Free) Shimmer and other artifacts: - Use iZotope RX 11 to filter out shimmer: de-click (single-band), de-crackle (high), de-reverb (adaptive)