While I understand the curation challenges would be non-trivial, the very best audio solution would be to permit authors to record their own audio tracks of their stories. Automated curation could be done these days by training AI on each author's voice and then ensuring the voice track matches the written word, which is already auto-curated by AI. Ultimately, Medium is about connecting content providers with content consumers, and the author's own voice is infinitely better at accomplishing the task than any machine-generated impersonal "voice."