Universo Punzadas
Punzadas Sonoras is my favourite podcast: I'm fascinated by how Inés and Paula choose their subjects and the way they get into them. I always wanted a library of the works, authors and references from their 100-plus episodes, so I built one.
The question
Official episode descriptions only tell a fraction of what's actually cited. What if you could search any author, work or subject mentioned on the podcast, with the exact quote and the minute it was said?
Universo Punzadas was born from that question. It catalogues every cultural reference cited across the podcast's 118 episodes: books, films, records, authors, and the subjects that connect them.
How it works
Every episode is transcribed and read in full, word by word. Only what can be pinned to a literal quote in the audio makes it into the catalogue — nothing is filled in from outside knowledge.
Transcripts are generated with mlx-whisper (large-v3 model) and pyannote 3.1 diarisation, running locally. Extracting references — what's cited, who cites it and in which episode — is done by hand, reading the full transcript against a documented editorial criterion.
Explore the catalogue
Search it directly here: by author, work, subject or episode, with charts that connect the references to each other.
If it doesn't load above, you can open it in a new tab.
Why this project matters to me
Universo Punzadas brings together two things I love: listening to the podcast and building data tools. It's a project that demands patience — hours of audio listened to and transcribed — and editorial care: every data point has to be traceable to a literal quote.
To me it represents what I love most about working with data: turning something scattered and hard to find into something navigable, searchable and shared.
Technology stack
From voice to catalogue
Every new episode goes through the same process, start to finish:
Explore the code
Universo Punzadas is open source. Explore the code or visit the full catalogue.