I want a “record voice” feature that automatically uses local models like on-device Whisper or Parakeet, so the product keeps getting better without sacrificing privacy or speed. Add more cloud choices, like Soniox, Groq, and Deepgram, so people can switch providers easily and still get fast, high-quality transcription. With more AI cloud providers coming up, I think moving between them should feel clearer and easier, guided by obvious on-screen cues, not just a dropdown menu.