Use ElevenLabs’ cloning feature but follow consent and audio best practices: obtain written consent covering reuse/commercial use; record 20–60 minutes of clean, dry speech in a treated room as 24–48 kHz WAV; vary pitch/intensity with multiple scripts; trim breaths/clicks and keep labeled takes; run a small pilot clone, fix artifact sections by re-recording or light denoising, then lock voice usage and metadata. Use ChatGPT to generate consistent ad/intro scripts for testing.