Audio (TTS & STT)

Yunxin mirrors the OpenAI Audio API for text-to-speech and speech-to-text, plus a voice-management surface for listing, cloning, and designing voices. Discover audio models with the Models API (GET /v1/models?type=audio_speech or type=audio_transcription).

Text-to-speech (TTS)

POST /v1/audio/speech

Synthesizes input text into an audio stream. The response body is the raw audio file (not JSON).

Parameters

Example

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["YUNXIN_API_KEY"],
    base_url="https://api.yuhuanstudio.com/v1",
)

response = client.audio.speech.create(
    model="model-id",
    voice="nova",
    input="Welcome to Yunxin, your unified AI API gateway.",
)

response.stream_to_file("output.mp3")

Speech-to-text (STT)

POST /v1/audio/transcriptions

Transcribes audio in its original language. Send the audio file and model as multipart form data.

Parameters

Example

with open("recording.mp3", "rb") as audio_file:
    transcript = client.audio.transcriptions.create(
        model="model-id",
        file=audio_file,
        language="en",
    )

print(transcript.text)

Translation

POST /v1/audio/translations

Transcribes and translates audio into English. Same multipart shape as transcription, minus language:

with open("chinese_audio.mp3", "rb") as audio_file:
    translation = client.audio.translations.create(
        model="model-id",
        file=audio_file,
    )

print(translation.text)  # English translation

Voices

GET /v1/voice

Lists the voices available for a provider (provider query parameter, default openai). Each entry has an id, name, and type (preset, cloned, or designed). Providers without a native voice catalog fall back to the standard OpenAI preset voices:

curl "https://api.yuhuanstudio.com/v1/voice?provider=openai" \
  -H "Authorization: Bearer $YUNXIN_API_KEY"

Clone a voice

POST /v1/voice/clone

Creates a custom voice from a short audio sample (provider-dependent; e.g. MiniMax).

Design a voice

POST /v1/voice/design

Generates a brand-new voice from a text description.

Both endpoints return a record with id, type, and a status (pending, processing, completed, failed). Use the returned voice id as the voice value in TTS.