We now have a benchmark for how often we're tricked into thinking a voice is human. Here is how ...
A new benchmark measures how often humans are tricked into mistaking synthetic voices for real ones by asking participants to identify the human voice between two audio recordings of the same text. The author notes that Gemini 3.8 Flash TTS is currently the best model at fooling people. Open article →
- Drop 2/15 <br /> <br /> Are we actually measuring the things we ultimately want our voice AI models to achieve?<br /> <br /> The best AI voices are becoming hard to tell apart.<br /> From each other. Not from humans.<br /> <br /> 🚀 Introducing Turing TTS: an evaluation that measures how often human listeners mistake AI-generated speech for real human speech.<br /> <br /> We tested leading TTS models against real human speech. Even the best model fooled listeners only ~10% of the time.<br /> <br
- The best AI voices sound like each other now. Not like us. @shobhitbanga and @voicearena_ai jus...
- Drop two from Voice Arena is a test called Turing TTS. TTS is text-to-speech: the model that give...