Spotify is testing a way to make some podcasts available in other languages while keeping the sound of the original speaker. Its Voice Translation pilot uses OpenAI voice synthesis to translate English voices into Spanish, French, and German. The approach could make translated episodes feel closer to the original conversation, while also making mistakes harder for listeners to recognize.
A familiar voice in another language
The pilot is limited to select podcasters, including Dax Shepard, Monica Padman, Lex Fridman, Bill Simmons, and Steven Bartlett. Spotify says translated episodes will be available worldwide to both Premium and Free users. Listeners can find supported translations in the “Now Playing View” or through a dedicated Voice Translations Hub, which Spotify says will add more translated content.
Keeping a speaker’s distinctive vocal characteristics is a central part of the idea. Traditional dubbing can replace a speaker’s voice with another performer’s; Spotify’s system instead aims to make the translation sound like the podcaster. That may help listeners feel a closer connection to the person speaking, even when they do not understand the original language.
Lex Fridman shared a Spanish sample of his cloned voice on X. He wrote, “This is me speaking Spanish, thanks to amazing work by Spotify AI engineers. The translation & voice-cloning are fully done by AI. Language can create barriers of understanding & thus fuel division. I can’t wait for AI to break down this barrier & reveal our common humanity.”
Translation adds a second layer of uncertainty
Voice synthesis and translation each bring their own challenges. Voice-cloning systems analyze source audio and use training data to create a similar voice. The results can be less reliable when a person’s vocal style, including certain accents, is not well represented in the training samples.
Spotify’s pilot combines that kind of voice generation with the task of carrying meaning from one language to another. AI translation has improved over the past decade, but it can still miss nuance and cultural context. Those gaps matter in podcasts, where meaning may depend on a turn of phrase, a joke, or details from a longer conversation.
A listener who knows the translated language may notice when wording sounds wrong. But a listener who does not know it may have little way to check. The original speaker could face the same problem: if they do not speak the translated language, they may be unable to tell whether the audio preserves their intended meaning.
When the translation sounds like the speaker
A machine translation can be framed as such, setting an expectation that the wording may not be perfect. A translation delivered in the speaker’s own voice can blur that distinction. If a translated clip is shared without its context, someone could mistake it for something the podcaster originally said in that language.
That creates a reputational concern as well as a quality concern. The speaker’s voice gives the translation a personal feel, but may also make listeners more likely to associate an error with that person. The risk is particularly difficult to manage when the speaker cannot review the translation for themselves.
Reactions to the pilot have not been uniformly positive. Jeremy Parish, co-creator and co-host of Retronauts, wrote on BlueSky, “Another reason to roll my eyes when people ask why we don’t make the podcast available on Spotify.” His response points to a broader question about creator comfort with AI-generated versions of their work.
A limited pilot leaves important questions open
Spotify’s program currently operates on a limited, opt-in basis with selected podcasters. In that setting, the source article says concerns about cloning podcast guests’ voices do not appear to be at play. A wider rollout could make consent more complicated, especially if episodes include people who did not agree to have their voices translated and synthesized.
Spotify says it plans to gather feedback from creators and listeners as it refines Voice Translation. That feedback may help show whether the feature is useful, whether translations communicate what speakers intended, and how clearly listeners understand that the audio is AI-generated.
The pilot offers a possible route for podcasts to cross language barriers without replacing the familiar voice at the center of an episode. Its success will depend on more than how natural the voice sounds. Listeners and creators also need confidence that the translated words preserve the original meaning and remain identifiable as a translation.