podcast microphone

Case

Svenska Dagbladet

Media & Marketing

Unlocking podcast insights and making them searchable

Speech-to-textNLPProcess efficiencyNew business

Modulai worked with Svenska Dagbladet to build a tool for transcribing and analyzing Swedish podcasts. It lets journalists explore audio content through a single interface, surfacing topics, named entities, emotional tone, and linguistic detail across episodes.

Outcome

  • 696

    Hours of audio transcribed and analyzed

  • 548

    Episodes across more than ten years

  • <1

    Minute o search a decade of audio

    • Challenge

      As podcasting has grown, manually keeping track of what is said across hours of audio has become impractical. Svenska Dagbladet wanted a faster way for journalists to find and understand what was discussed in a podcast, without listening through every episode.

    • Solution

      Modulai combined speech recognition with natural language processing to transcribe podcasts and make them searchable. Whisper handles the transcription, and a set of NLP models analyzes the text for topics, entities, and tone. A custom interface lets journalists search and explore the results directly, turning hours of audio into something they can query in seconds.

    • Tools

      The project used OpenAI's Whisper for speech recognition and pre-trained NLP models for the linguistic analysis. Computational stylistics were applied to measure language complexity. A search interface was built on top, so journalists could explore the content themselves without technical support.

    • Value created

      Instead of scrubbing through audio to find a quote or check who said what, journalists can query transcripts directly and see the topics, people, and tone across an episode. That makes it to work with hundreds of hours of audio at once, far more audio than manual listening would allow.

    Learn more

    Related content