From article to podcast: how do we work?

Preparing and enriching the content

A written article and a podcast are not built the same way. Before audio generation, the content is therefore prepared for its adaptation to this new format.

Where the subject calls for it, the working basis can also be enriched with references or additional documentary material. These add context, clarify certain historical or scientific facts, or provide what is needed for a better spoken understanding of the subject.

The goal is to have material that is solid and well documented enough before moving on to producing the podcast.

Using AI to move from writing to audio

This working basis, together with editorial instructions, is then submitted to an artificial intelligence tool specialised in creating audio content.

This is the step where AI fully plays its part: turning largely written material into an experience designed for listening.

It is not simply a matter of reading the article aloud. The podcast format requires adaptation: rephrasing certain explanations, introducing transitions, restoring context that is no longer visible on screen, and building an exchange natural enough to hold the listener's attention.

Here, AI is a production and multimedia adaptation tool. It gives a new form to Tangente editorial content that until now existed mainly in written form.

Then comes Tangente's review

Generating the audio file is not the end of the process.

Before publication, every podcast is listened to and assessed by the Tangente team.

This step is essential, because generative artificial intelligence technologies can produce approximations, introduce errors, or misinterpret parts of their source material.

Our validation covers faithfulness to the content and spirit of the original article, the accuracy of the mathematical elements, and that of the historical and scientific facts mentioned in the podcast.

We also pay attention to the quality of the popularisation. The podcast must stay understandable and instructive as far as the nature and complexity of the subject allow, without simplifying to the point of distorting the concepts covered.

Finally, we assess the podcast as an audio work in its own right: clarity of the explanations, overall coherence, pacing, and the general quality of the listening experience.

Only once this review and validation step is complete is the podcast published on the Tangente Podcasts platform.

AI as a tool, Tangente as the publisher

Our approach rests on a simple distinction: artificial intelligence takes part in making the podcast; Tangente retains scientific and editorial direction of it.

We want to benefit from what these new technologies make possible today — in particular to make mathematical culture accessible in more forms — without giving up the safeguards that publishing scientific content requires.

The experience of the Tangente editorial team, its knowledge of mathematics, and several decades devoted to popularising it are the framework within which these new tools are used.

AI lets us explore a new way of passing knowledge on. It replaces neither the editorial work, nor the critical eye, nor the responsibility of the editorial team.

Spotted an error? Tell us

Despite the care taken in preparing, listening to and validating each podcast, some errors or inaccuracies can, as in any publication, escape the editorial team's attention.

If you spot a mathematical, historical or factual error, or any other inaccuracy in one of our podcasts, please do tell us.

Your feedback helps us maintain the editorial and scientific standards Tangente is committed to and, where necessary, correct or clarify published content.