Chinese text-to-speech

Voices and samples

Every sample is this sentence, generated through the API with default settings: “今天天气很好，下午我打算和朋友一起去公园散步。”

Round-trip check: we transcribed each sample with Whisper Large v3 and compared it with the text we sent. It tests whether the speech is intelligible to a machine, not whether it sounds natural to you; for that, listen. Traditional and Simplified forms of the same character were counted as the same.

Quickstart

curl https://api.inferenceapis.com/v1/audio/speech \
  -H "Authorization: Bearer $INFERENCE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "hexgrad/Kokoro-82M", "voice": "zf_xiaobei", "input": "今天天气很好，下午我打算和朋友一起去公园散步。", "response_format": "mp3"}' \
  --output chinese.mp3

import os
from openai import OpenAI

client = OpenAI(base_url="https://api.inferenceapis.com/v1", api_key=os.environ["INFERENCE_API_KEY"])
audio = client.audio.speech.create(model="hexgrad/Kokoro-82M", voice="zf_xiaobei", input="今天天气很好，下午我打算和朋友一起去公园散步。")
audio.write_to_file("chinese.mp3")

Notes for Chinese

• The voices are Mandarin. There is no Cantonese voice.
• Numbers, dates and abbreviations are read as written. Spell out anything that must be pronounced a specific way.
• Pacing follows punctuation; there is no SSML. speed changes the overall rate.

Cost and limits

Kokoro is billed at $5.20 per million Unicode characters of input. The 23-character sample above is 5.4 seconds of audio and costs $0.00012; at that rate an hour of Chinese speech is about 15,000 characters, or roughly $0.08. Each character counts once, however many bytes it takes to encode. A request accepts up to 20,000 characters; for longer text, split on sentence boundaries and join the audio. A paragraph returns in under a second in our measurement. Orpheus is English-only, so Chinese always uses Kokoro.

Other languages: English · Japanese · Spanish · French · Hindi · Italian · Portuguese

Speech-to-text in Chinese

The reverse direction: Whisper Large v3 transcribed our Chinese test clips with 9.5% character error. See Chinese speech-to-text.