AI Technology FAQ

Is OpenAI used for speech recognition? Is Whisper API used?

We use our own Automated Speech Recognition (ASR) technology, which has a higher accuracy rate than Whisper's if you compare them.

Is GPT used for Q&A? Is it based on ChatGPT? Is it GPT3.5 or GPT4? Is the Q&A model self-developed?

The Q&A part of the robot is mostly based on OpenAI's GPT-3.5 (ChatGPT) at present, and we plan to integrate GPT-4 into some advanced features in the future. Meanwhile, our self-developed LLM (Large Language Model) is currently in the experimental stage and will be launched soon.

What model is used for TTS? What are the features of MyShell's TTS?

We use our own TTS (text-to-speech) model, and the English TTS currently supports fast voice cloning, which can clone anyone's voice with only 1-5 minutes of voice samples.

MyShell's TTS has the feature of being closer to human pronunciation than other products on the market, as well as faster response time and lower computing costs. Currently, the technology for different emotional voices is in the experimental stage and will be launched soon.