Skip to main content

Text-to-Speech / TTS

Convert text to audio files with streaming support. Uses the /v1/audio/speech endpoint, compatible with the OpenAI API.

Basic Usage​

from pathlib import Path
import openai

client = openai.OpenAI(
    api_key="your-api-key",
    base_url="https://live-turing.cn.llm.tcljd.com/api/v1",
)

speech_file = Path("hello.mp3")

response = client.audio.speech.create(
    model="turing/tts-1",
    voice="alloy",
    input="Hello, the weather is nice today.",
)
response.stream_to_file(speech_file)
curl $TURING_BASE_URL/audio/speech \
  -H "Authorization: Bearer $TURING_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "turing/tts-1",
    "voice": "alloy",
    "input": "Hello, the weather is nice today."
  }' \
  --output hello.mp3

Parameters​

ParameterDescriptionValues
modelTTS modelSee Audio Model List; example uses turing/tts-1
voiceVoiceOpenAI: alloy / ash / ballad / coral / echo / sage / shimmer / verse; Qwen: Cherry / Serena, etc.
inputText to synthesizeUTF-8 string
response_formatOutput formatmp3 (default) / wav / flac / opus / pcm / aac
speedSpeech speed0.25 – 4.0, default 1.0

Full schema: /api/create-speech

Voice Quick Reference​

OpenAI voices (more natural emotional expression): alloy, ash, ballad, coral, sage, verse (recommended) / echo, shimmer, alloy (alternatives).

Qwen Omni voices (more natural for Chinese): Cherry, Serena, Ethan, Dylan, etc. (see the Alibaba Cloud DashScope documentation for the full list).

Streaming Response​

stream_to_file is a convenience wrapper; the underlying transport is chunked HTTP. To consume the stream directly:

with client.audio.speech.with_streaming_response.create(
    model="turing/tts-1",
    voice="alloy",
    input="...",
) as resp:
    with open("out.mp3", "wb") as f:
        for chunk in resp.iter_bytes():
            f.write(chunk)

Billing​

Billed by the number of input characters (see Billing & Usage for per-provider pricing). Speech speed, output format, and voice selection do not affect the unit price.

FAQ​

  • Qwen voice throws an error with the OpenAI SDK → Set model to the Qwen TTS model name; the voice name must also match the model.
  • Output contains noise → Switch to response_format: "wav" or use a higher-quality TTS model from the Audio Model List.
  • Long text is truncated → Synthesize in segments and concatenate; keep a single input under 4096 characters.
  • HTTPS stream drops → The client timeout is too short; increase timeout or split the input into segments.

See also​