OpenAI 本周五展示了其名为 “Voice Engine”（语音引擎）的预览版本。只需提供 15 秒的音频样本及文本提示，该引擎即可提供声情并茂的高仿真语音，令人难辨真假。OpenAI 正在测试该语音引擎，连测试人员都要签署一份防止滥用的声明。OpenAI 不想立即公开该模型，因为担心它会被不法分子滥用，特别是正值美国大选即将展开之际。该工具未来可能将收费。Meta 去年也发布了类似的工具 Voicebox，但该公司也没有开放该工具。
OpenAI showcased a preview version of its “Voice Engine” this Friday. With just a 15-second audio sample and a text prompt, this engine is capable of delivering highly realistic and emotionally expressive voices. OpenAI is currently testing the voice engine and even requires testers to sign a declaration to prevent misuse. OpenAI is hesitant to release the model immediately due to concerns about potential abuse by malicious individuals, especially with the upcoming US elections. It is possible that this tool will be offered as a paid service in the future. Meta also released a similar tool called Voicebox last year, but the company has not made it openly available either.
微信扫一扫