This story is part of the series from the Transition Accelerator 2024 cohort.
Meet Capital, Bulgaria’s largest business media, offers their readers in-depth analyses of current events, helping them make informed decisions to develop their projects confidently. Recognizing the potential of audio to seamlessly fit into their readers’ busy lifestyle, Capital embarked on a journey to develop an AI-enhanced audio feature for their articles.
IPI’s Media Innovation team talked to Dobromir Kotsev, Product Manager and Zornitsa Stoilova, Features Editor and Newsroom Development Lead of Capital, Economedia AD, about their AI strategy and lessons from the audience research that has helped their product resonate.
Initial feedback fuels a user-centric approach
Before the Accelerator, Capital had already experimented with a text-to-speech player on their website. However, when surveyed for feedback, users were not satisfied: the pronunciation in Bulgarian was inaccurate, and the AI-voice sounded too synthetic. Despite these shortcomings, data indicated that users who listened to articles were two times more engaged than typical users and were more likely to consume additional content. Improving the listening feature thereby proved a critical step in bolstering Capital’s user journey and subscription funnel.
Adopting learnings from the Transition Accelerator bootcamp in Vienna, the team conducted empathy interviews with users who took the initial feedback survey. Open-ended questions on lifestyle and news consumption led to insightful findings on audio content use: for example, many like to listen to articles at a very high speed, as they try to squeeze content consumption into their busy routines.
“We really tried to approach these interviews like a discovery process – the learnings not only helped us with this audio project, but also let us better place Capital’s content in our users’ daily life overall”, explains Dobromir Kotsev, Product Development Manager at Capital.
Rigorous testing and refinement of cloned voices
After shortlisting two voice cloning vendors which support the Bulgarian language, the team ran a lengthy and intensive testing process with both providers. They created separate datasets of 15 different AI-audio articles and had volunteer testers – mostly existing Capital subscribers – grade the cloned voices. Throughout this process, the team maintained close communication with the testers, emphasizing the direct impact of their input on the product they were paying for.
Capital’s main goal was to achieve a cloned voice that sounds as natural and calm as possible, so as not to create any distraction from the content and offer users a more pleasant and engaging listening experience. The vendor the team settled with still had flaws regarding pronunciation, but was willing to continue to collaborate to improve the quality of the product over time.
Capital’s audio UX website integration
Boosting engagement and subscriptions
Better voice quality ultimately unlocked avenues for partnership and monetization for Capital: they are currently launching a shared subscription-bundle programme with Storytel, a popular audiobook provider in Bulgaria (new Capital subscribers get a Storytel subscription and vice versa). Furthermore, Capital’s audio products will be available on Storytel’s platform, creating a direct revenue stream.
Re-working their audio strategy also enabled Capital to dig deeper into how to make listeners stick with the content: They are in the process of rolling out a feature for curated playlists, where thematically connected articles play one after another, encouraging continuous listening and increasing the chances of listeners turning into paid subscribers.
The Transition Accelerator is part of the Media Innovation Europe programme (MIE), made possible with the support of the European Union.
