How can news be processed quickly, automatically and without manual work? At YouRadio Talk we launched a format that summarises the key topics from Czech news podcasts every day. Transcription, summarisation and the voice of the robot Anet all run fully automatically. This is what development looks like when technology, product and content work together.
Summaries built on data
Every day we create a digest of the most important points from the current episodes of the most listened-to news podcasts, and we automatically select the topics that resonate most on that day. We use speech-to-text technology, which lets us break podcasts down into precisely structured content. From the text we derive summaries, keywords, main topics and headlines.
Language models then process this foundation and automatically compile the daily news overview. Their summary captures what was discussed and what conclusions were reached, and under each episode you can find where the information comes from.
This text is then turned into a five-minute audio episode voiced by Anet, a voice AI built on ElevenLabs technology. The episode description also includes links to the individual podcasts, so if a topic catches your interest, you can play the full episode right away.
No manual steps, full automation
The whole process is fully automated, from transcription and content selection to creating the episode and publishing it. Ondřej Musil as project manager and Rudolf Grabowski, a developer specialising in working with language models, developed the format. Thanks to the combination of development, data and product, the format delivers current content in the required quality every day without intervention.
Your 5 minutes a day
You can listen to Anet's news roundup in the app and on the YouRadio Talk website. If you follow the news category, the episode will automatically appear first in your timeline. You can listen to the five-minute news, without having to switch between sources, in the YouRadio Talk news roundup.
Comments