We are featuring the products that won the Radio World Best of Show Award at the 2026 NAB Show. Winners were chosen by a panel of engineers and RW editorial staff for their innovation, feature set, cost efficiency and performance in serving the industry. Information about how the program works can be found in a FAQ here.
“In live broadcast, a single AI hallucination isn’t just a glitch, it’s a compliance nightmare,” says MusicMaster.
“While the industry rushes to adopt Large Language Models, the missing link has been a ‘trust layer’ that bridges generative power with broadcast-grade accuracy.”
A partnership of MusicMaster and musiccompanion from SwissMediaPartners introduces a human-verified data layer between AI systems and broadcast output. Its core concept is a “song molecule,” a structured, curated unit of information that the company says transforms unstructured AI noise into a deterministic data asset.
“These contain verified metadata and contextual information sourced from authoritative references and reviewed by a designated editor before being made available to AI processes.”
Integrated tools assist the user in comparing and selecting reliable metadata sources. This ensures that high-quality data is used to enrich a music library, whether for scheduling, reporting or presentation. AI-assisted song analysis supports coding and classification tasks from within the same framework. A key principle is that the AI is not permitted to operate on the open internet or with unverified data.
This architecture also supports AI-assisted voicetracking. Script generation is constrained to verified molecule data, and its prompting strategies produce broadcast-appropriate language that aligns with your station tone and format. These scripts can be rendered as audio using synthetic voices and delivered into playout.
Info: www.deliver.media/musiccompanion
