A developer going by the handle stoilms has released a companion eBook for what might be the most oddly specific audiobook project of the year—an AI-generated narration of Homer's The Odyssey using a cloned voice of Sir Michael Caine, powered by ElevenLabs. The catch? There's a naming convention mismatch that's been bugging listeners who want to follow along with the text.

The Problem: Greek vs Roman Mythology Names

The audio narration uses Greek names throughout—Odysseus, Poseidon, Zeus—but the William Cullen Bryant translation that serves as the source text relies on Roman nomenclature. That means Ulysses instead of Odysseus, Neptune instead of Poseidon, and Jove in place of Jupiter. When you're listening to one set of names while reading another, following the epic's geography and mythology becomes unnecessarily tedious.

The Solution: A Matching eBook

Rather than waiting for ElevenLabs or whoever commissioned the audiobook to release an updated version, stoilms took matters into their own hands. The companion eBook aligns with the audio narration by using Greek names throughout while preserving Bryant's Victorian-era prose style. This lets listeners read and hear the same terminology simultaneously—a surprisingly important detail when you're deep in a twelve-hour journey through ancient Mediterranean waters.

Technical Implementation

The project leverages Python for text processing, handling the name substitutions systematically rather than doing find-and-replace manually across 12 books of epic poetry. The developer notes that maintaining consistency required tracking context—certain passages needed different handling based on whether characters were speaking in dialogue or narration. The eBook format chosen allows readers to bookmark positions and sync with audio timestamps.

Open Source Approach

By releasing the project on GitHub, stoilms enables other developers facing similar translation-audio mismatches to adapt the approach for their own audiobook projects. This is exactly the kind of grassroots tooling that emerges when AI-generated content meets real-world usability gaps—someone identifies a friction point and builds a fix without waiting for corporate solutions.

Key Takeaways

  • AI voice cloning enables celebrity-narrated audiobooks from dead or retired performers, but source text selection requires careful matching to narration choices
  • Name consistency between listening and reading is critical for comprehension in mythology-heavy texts like The Odyssey
  • Companion tools built by communities often outpace official releases when naming or formatting mismatches emerge

The Bottom Line

This project exemplifies why the best AI tooling often comes from users solving their own problems rather than waiting for polished enterprise solutions. Stoilms saw a friction point, identified it was purely a naming convention issue, and shipped an open-source fix in less time than it takes to complain about the mismatch on social media. That's hacker culture at its finest.