After HeyGen, this dubbing technology which makes the presenter speak in several languages, by synchronizing lip movements, Spotify has just announced a comparable innovation for podcasts.
Spotify is launching into automatic translation of podcast voices, using artificial intelligence, with one objective: the whole world. Let’s take one of the first five examples uploaded by the Swedish streaming service. This is one of the most listened to podcasts in the United States, in the “Technology” category: the podcast by Lex Fridman, himself a computer scientist and artificial intelligence researcher. This Ukrainian of origin, who today has dual American and Russian nationality, first worked at Google, and he now teaches at the prestigious MIT near Boston.
Same episode, same voices, several languages
In the original version, in other words in English, he speaks in his voice, well known to fans: a deep and calm voice, quite easily recognizable. Until now, this was the only version available. But episode 390 is now available, for all Spotify users, in a second version, in Spanish, generated using artificial intelligence and the latest voice technologies from OpenAI, the creator of chatGPT.
The feat is that it is not a synthetic, robotic, random voice, but the cloned voice of Lex Fridman, in a language he does not master: Spanish. By clicking on Play, we discover the same beginning of the episode with the same voice… But in Spanish.
Lex Fridman, on Instagram, was happy to suddenly speak Spanish: “I can’t wait for AI to break down this language barrier, he wrote, to reveal this humanity that we have in common”. On Spotify, you can listen to this entire episode, which lasts almost three hours. Everything is there: the tone, the silences, the rhythm. Not bad for a pilot version!
In French and German, soon
And the competitor of Apple Podcasts, Deezer and Amazon Music does not intend to stop with Spanish. The translations, this time in French and German, will arrive in the coming days and weeks. This is, in any case, what Spotify promises to its half a billion users in total, including 210 million paying subscribers in the last quarter. There are now more than 100 million people listening to podcasts on the platform.
So, no false joy: only a few podcasts will benefit from these translated versions, at least initially. And then, the other challenge is the translation of podcasts from languages other than English. Imagine if French podcasts could become available, with the original voices, in English, Spanish or Chinese. They could go beyond the borders of the French-speaking world, with new perspectives.
The stakes are high: the competition is certainly not going to stop there, in particular Apple Podcasts which is struggling to defend its No. 1 position after inventing podcasts more than 15 years ago.