@Dululu For Chinese voice AI using Vapi.ai, the best voice model depends on whether you're prioritizing TTS (text-to-speech) quality, STT (speech-to-text) accuracy, latency, or cost. Here's a breakdown of the best current options for Chinese support in Vapi-compatible models: --- 1. ElevenLabs (for TTS) Pros: Very high-quality, expressive, near-human voices. Chinese Support: Limited. Currently, it's experimental and not fully natural for Mandarin. Use if: You prioritize ultra-realistic voice quality for English but not ideal for Chinese yet. --- 2. OpenAI TTS (via Vapi) Voices: onyx, nova, echo, etc. Chinese Support: Good, works well with Mandarin, especially if the input is clean. Pros: Low latency, good tone, natural pacing. Use if: You want smooth, multi-language support including Chinese, and high-quality TTS. --- 3. Azure Neural TTS (Microsoft) Voices: zh-CN-XiaoxiaoNeural, zh-CN-YunjianNeural, etc. Chinese Support: Excellent, with regional accents and expressive style options. Pros: Highly realistic, supports SSML, stable. Use if: You want the best Chinese TTS, especially for a production use case. --- 4. Google Cloud TTS Voices: cmn-CN-Wavenet-A, cmn-CN-Wavenet-D, etc. Chinese Support: Solid, reliable. Pros: Good quality, decent latency, supports various dialects. Use if: You're already in the Google ecosystem and need solid TTS. --- 5. Deepgram (for STT - speech-to-text) Language: Mandarin (zh) Accuracy: Very good, especially with clean audio. Use if: You want real-time transcription of Chinese voice input. Works well with Vapi. --- Recommended Pair for Chinese Vapi Bot: TTS: Azure zh-CN-XiaoxiaoNeural or OpenAI onyx STT: Deepgram with language set to zh