HarperCollins turns to ElevenLabs for foreign-language audiobooks

Audiobooks are notoriously expensive to make. Between studio time, directors, and professional voice talent, producing dozens of hours of recorded audio means publishers rarely bother recording anything outside of proven bestsellers.

Publishing giant HarperCollins is now tapping AI to tackle the dusty corners of its catalog. The publisher announced a partnership with ElevenLabs to generate audiobooks for its foreign-language markets, specifically focusing on deep backlist series that otherwise would never have received audio adaptations.

Breathing synthetic life into deep backlists

The initiative relies on ElevenLabs’ Studio tool, a production suite aimed at publishers and authors. Through its Projects interface, editors can select or design custom voices, assign specific lines to different speakers, and tweak pauses while keeping contextual tone coherent.

According to ElevenLabs, the workflow can piece together a full audiobook in as little as one hour. That dramatically undercuts the steep operational overhead of a conventional studio setup.

Still, HarperCollins was careful not to antagonize the creative community. The publisher explicitly stated that voice-actor-led productions remain intrinsic to its core audiobook strategy. ElevenLabs’ text-to-speech engine is being framed purely as a complementary tool to bring obscure non-English catalog titles to market.

ElevenLabs CEO Mati Staniszewski noted that without AI, the industry simply lacks the time and money to unlock these older works for international readers.

Why it matters

This is a remarkably smart strategic wedge for HarperCollins and generative voice tech.

By restricting AI strictly to backlist titles in non-English markets, HarperCollins largely skirts the immediate backlash of displacing frontline voice talent. They aren't automating blockbuster audiobooks; they are creating audio products out of dead inventory where unit economics previously forbade it. For long-tail authors, it turns dormant rights into passive royalties.

For the publishing sector, this legitimizes generative audio at the enterprise scale. Once a Big Five publisher proves that AI-generated audio is commercially viable for backlists, rival houses will face immense pressure to follow suit.

However, there is an obvious tension here. While publishers pitch AI as a mere supplement today, the economic gap between an hour of compute and weeks in a recording booth is massive. The long-term challenge will be whether the boundary between human-narrated prestige titles and automated catalog filler holds once listeners stop noticing the difference.