‘Eurostars’ approves 1 million funding for AI-audiobooks

The AAI Labs-led international consortium has been selected to obtain public funding in the recent ‘Eurostars’ call. The partners aim to create a comprehensive AI-driven audiobook creation suite, where publishers will be able to produce audiobooks cheaper while retaining the high quality of the narrating voice.

The consortium team is composed of a machine learning company AAI Labs (Lithuania), an international award-winning publishing house in South Africa Quickfox Publishing and a leading Croatian new media and e-books publisher Bulaja Naklada. 

The project will develop a low-cost, multi-language, high-quality semi-automated audiobook creation system based on extremely precise text-to-speech (TTS) AI synthesis algorithms. Partners plan to research on creating a TTS system that would allow transferring voice features (children's voices, foreign accents) from language to language, introducing a more theatricalised user experience.

Globally, audiobooks have around 6% of the total book market share which is growing yearly as more and more people are preferring to listen to books. But the majority of written works are not accessible in audio format due to the high cost of narration and production.

“The market for audiobooks is enormous given the prevalence of smart devices with streaming capabilities. A recent breakthrough in natural language processing allows human quality-like synthetic speech. Cutting the audiobook production cost by this method and popularising literature from lesser-known publishers and authors will benefit the population, encouraging more people to read through listening,” says Aistis Raudys, the CEO of AAI Labs.

The AI-based method that partners seek to develop includes only manuscript preparation, TTS technology application, and post-production mastering, thus making narration work and audio engineering superfluous. The overall production time will be cut down from weeks to as little as an hour, assuming little post-production is necessary.

The TTS technology will be mainly targeted to apply in areas where audiobooks are unprofitable to produce with traditional techniques. This will expand consumer access to a broader range of literature, including materials aimed at indigenous, minority groups, and niche markets, all creating a positive impact on society that is inclusive of everyone thanks to AI.

“Cheaper production means that publishers and educators can roll out a larger number of audiobooks in developing countries, where literacy is often low, but access to cellular devices and bandwidth is now readily available. This can be a game-changer in terms of reaching the masses in a language and format that they can easily access via devices they already own,” explains Vanessa Wilson, the publishing director at Quickfox Publishing.

This is particularly relevant in countries in Africa, where literacy rates are estimated at 67.4% as of 2021, yet mobile services' access is at around 650 million users, according to African Development Bank.

“Most of the research in TTS was done using 22050 Hz sampling rate recordings. Such quality is insufficient for most audiobook publishers as 44100 Hz - 48000 Hz sampling rate is the industry standard. We already have some past experience in successfully building TTS systems supporting such higher-quality audio. This gives us an upper hand with the audiobook publishing companies as the quality requirements will be met,” states Tadas Subonis, the CTO of AAI Labs.

The project is expected to end by 2025, with the audiobook creation suite then ready to purchase for publishing houses and other media companies.

‘Eurostars’ is the largest international funding programme for SMEs wishing to collaborate on R&D projects that create innovative products, processes or services for commercialisation, part of the European Partnership on Innovative SMEs. The partnership is co-funded by the European Union through Horizon Europe.