Search papers, labs, and topics across Lattice.
This study investigates how speech-to-speech (S2S) models handle gender attribution and voice rendering by examining the interplay between a speaker's voice and the gender stereotypes embedded in the content. Through a controlled experiment involving multiple languages and models, the researchers found that while the output voice remained consistent and free from stereotype drift, the models predominantly based gender attribution on the content rather than the voice. This led to significant misgendering rates, particularly when there was a mismatch between the voice and the content, highlighting a critical gap in current S2S systems' handling of gender identity.
Every speech-to-speech model misgenders speakers based on content, not voice, with misgendering rates soaring to 90% when voice and content clash.
Speech-to-speech (S2S) models now run inside dubbing, translation, and voice agents. Unlike text models, they hear the speaker's voice, which carries the speaker's gender. A faithful system should treat a speaker as who they sound like, not as whoever usually says what they said. Testing this is harder than it looks, since most S2S models answer in a single, fixed output voice, hard-coded so it cannot drift toward a stereotype. Checking the output voice comes back clean even when the model is biased. We therefore ask two questions. When a model re-speaks the input, does the stereotype in the words shift the perceived gender of the output voice (voice rendering)? And when the model states the speaker's gender, does it follow the voice or the content (gender attribution)? We answer both with one controlled experiment crossing male and female voices with masculine-, neutral-, and feminine-stereotyped passages, on five open- and closed-source models in English, Spanish, and Mandarin. The rendered voice shows no stereotype drift. But every model decides the speaker's gender from the content, not the voice. Making the content one step more feminine (masculine ->neutral ->feminine) multiplies the odds of a"female"judgment by 1.7-24. When the content clashes with the voice, the worst model misgenders the speaker in 90% of cases. When they agree, it misgenders in only 2%. The bias thus hides in gender attribution, where fixed-voice evaluation cannot see, and where audits must look as S2S systems increasingly speak for real people.
