Claude is indeed impressive, but he’s not a god. Plus, Anthropic’s investment in the voice domain is relatively less compared to others. So this blogger actually got it backward. Right now, GPT’s paid model and Gemini are both native voice modes (Audio In, Audio Out). Claude, on the other hand, is the only one among the three that still sticks to the three-stage pipeline of STT, TTT, and TTS. Also, there’s a bit of an issue with the causality here. Eleven Labs is indeed one of Claude’s suppliers, but Eleven Labs mainly focuses on TTS (text-to-speech), which is the “mouth.” The issue the blogger mentioned about detecting fast speech speed is actually the “ear” part, which is the job of STT. And STT isn’t the primary focus of Eleven Labs. In fact, traditional STT can add some markers to sentences. For example, if someone finishes a long string of words in 10 seconds or doesn’t pause in the middle of a complex sentence, STT can record these and conclude that the speech speed is too fast. If you really want to practice English, I’d still recommend using ChatGPT’s paid model. Even though its voice mode core might still be GPT-4o, the native conversational experience feels really good.