Posted 3/30/2024, 1:57:03 AM
OpenAI Unveils New AI That Clones Voices from Brief Samples, But Proceeds Cautiously Due to Deepfakes Risk
- OpenAI unveiled new "Voice Engine" that clones speech from just a 15-second audio sample
- Showcased use cases like giving voice back to a patient who lost speech due to a brain tumor
- Technology currently powers ChatGPT's text-to-speech voices and translations for HeyGen
- Being cautious about broader release due to risks of voice impersonation and "deepfakes"
- Also working on releasing GPT-5 this year, new video generator Sora, and secret project "Q*" focused on reasoning
