I let software run my station for 8 weeks, unattended. What I learned.

Yaniv Morozovsky, an experienced radio professional, recently experimented with fully automating one of his internet stations for eight weeks. The station, accessible at https://ystream.live, remains on air to this day. In this account, we will explore Morozovsky's insights for those considering automated radio stations.

One crucial aspect of radio automation is the fade timer. Automating stations often face the issue of songs fading at the same point, leading to a cold ending that sounds unnatural. To address this, Morozovsky analyzed each song's intro, end, vocal entry, and outro, ensuring crossfade points were set per song. Once implemented, listeners no longer noticed the absence of a human host.

Making an automated station sound like a DJ mainly involves perfecting crossfades. Morozovsky even cloned his own voice for the breaks, but the voice itself received less attention from listeners than the timing of the talk-up relative to the vocal. When the transition lands correctly, the station sounds like a traditional radio program.

Morozovsky also experimented with AI-driven DJ segments, which included reading the weather and a short news segment at the start of each hour. He discovered that scheduling by rules, rather than clock-based schedules, yielded better results. Features like starting every hour with an international song, managing key, tempo, energy, era, genre, and artist separation, and applying written rules were instrumental in streamlining the process.

However, some challenges remained. The automation software sometimes repeated songs or played the same artist multiple times within an hour, creating annoying bugs. Once, a missing forecast led to a placeholder line airing on the station. Additionally, song titles containing symbols, such as "Cloud #9," were misread by the AI DJ, becoming "Cloud Nine" instead. Furthermore, the hour often landed within half a song of the :00 mark, rather than precisely on it, due to Morozovsky's decision to prevent songs from being cut.

Despite these hurdles, Morozovsky believes that automating small stations is the future. He built yStream Studio, his own software product, after witnessing the benefits of automation firsthand. He is open to answering questions about the segue analysis, voice selection, or any other aspect of his experience.