Voice Isolation From Background Music

I’m working with a high-quality audio recording that runs a 50 minutes. I need the spoken voice separated cleanly from the underlying music and returned to me as an MP3.

You’re free to use any technique you trust—spectral editing in iZotope RX, AI stem-separation tools such as LALAL.ai/Spleeter, or manual work in Adobe Audition—as long as the end result is crisp, intelligible speech with minimal musical bleed and no timing shifts.

Deliverables
• Voice-only stem, MP3 (320 kbps preferred), exact length match to source
• (Optional but welcomed) music-only stem in MP3 for reference

Acceptance criteria
• Background music residual at least 25 dB below speech throughout
• No new artifacts, phasing, or noticeable EQ coloration
• File names clearly identify each stem

If you can turn this around quickly and demonstrate similar past work, let’s get started right away.