Create speech ​

POST https://openqiao.com/api/v1/audio/speech

The Create speech endpoint converts input text into audio (Text-to-Speech). It is compatible with OpenAI's Create speech API.

By default this endpoint returns a single JSON response containing the base64-encoded audio and token usage. Set stream to true to receive the audio incrementally as Server-Sent Events instead. The audio encoding is controlled by the response_format parameter.

Request headers ​

Authorization string * ​

Bearer Token authentication, formatted as Bearer $OPENQIAO_API_KEY.

Content-Type string * ​

Request content type, defaults to application/json.

Request ​

model string * ​

The model ID to use, in the format <provider>/<model-name>, for example google/gemini-3.1-flash-tts-preview.

input string * ​

The text to synthesize into speech. Cannot be empty.

voice string Optional ​

The provider-specific voice identifier used to generate the audio. The set of valid values depends on the model.

response_format string Optional ​

The output audio encoding. Provider-specific; defaults to pcm. An unsupported value returns a 400 error.

• pcm — Raw PCM audio (signed 16-bit little-endian, mono, 24 kHz, no container).

stream boolean Optional ​

Whether to stream the audio. Defaults to false.

• false — Returns a single JSON response with the full base64 audio and token usage.
• true — Returns the audio as Server-Sent Events (Content-Type: text/event-stream), matching OpenAI's streaming shape. The audio is base64-encoded inside speech.audio.delta events (one per chunk, each carrying its mime_type), followed by a speech.audio.done event carrying token usage, then a final [DONE] event.

speed number Optional ​

Optional playback speed multiplier. Must be a positive number. Supported by select models only.

Response ​

Non-streaming (default) ​

On success, the response is a JSON object with the base64-encoded audio, its media type, and token usage:

{
  "model": "google/gemini-3.1-flash-tts-preview",
  "audio": "<base64-encoded audio bytes>",
  "mime_type": "audio/L16;rate=24000",
  "usage": {
    "input_tokens": 5,
    "output_tokens": 109,
    "total_tokens": 114
  }
}

• audio — base64-encoded raw audio in the requested response_format (PCM by default). Decode it to obtain the audio bytes.
• mime_type — the provider's media type for the audio bytes (例如 audio/L16;rate=24000), useful for wrapping or playback.
• usage — token usage with input_tokens, output_tokens, and total_tokens.

Streaming (stream: true) ​

The response is a text/event-stream of Server-Sent Events. Each event is a data: line containing a JSON object:

data: {"type":"speech.audio.delta","audio":"<base64-encoded audio bytes>","mime_type":"audio/L16;rate=24000"}

data: {"type":"speech.audio.done","usage":{"input_tokens":5,"output_tokens":109,"total_tokens":114}}

data: [DONE]

• speech.audio.delta — carries a base64-encoded audio chunk plus its mime_type (the provider's media type for that chunk). Concatenating and decoding all deltas yields the same bytes as the non-streaming audio field.
• speech.audio.done — carries the token usage object with input_tokens, output_tokens, and total_tokens.
• [DONE] — terminates the stream.

On error, the endpoint returns a JSON object instead of audio:

{
  "error": {
    "code": "...",
    "type": "...",
    "message": "..."
  }
}

curl https://openqiao.com/api/v1/audio/speech \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENQIAO_API_KEY" \
  -d '{
    "model": "google/gemini-3.1-flash-tts-preview",
    "input": "开桥 is an LLM API aggregation service.",
    "voice": "Kore",
    "response_format": "pcm"
  }'

const response = await fetch("https://openqiao.com/api/v1/audio/speech", {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    Authorization: `Bearer ${process.env.OPENQIAO_API_KEY}`, 
  },
  body: JSON.stringify({
    model: "google/gemini-3.1-flash-tts-preview",
    input: "开桥 is an LLM API aggregation service.",
    voice: "Kore",
    response_format: "pcm",
  }),
});

const { audio } = await response.json();
// Decode the base64 audio into raw bytes.
const bytes = Buffer.from(audio, "base64");
await import("node:fs/promises").then(fs => fs.writeFile("speech.pcm", bytes));

import base64
import os
import requests

response = requests.post(
    "https://openqiao.com/api/v1/audio/speech",
    headers={
        "Content-Type": "application/json",
        "Authorization": f"Bearer {os.environ['OPENQIAO_API_KEY']}",  
    },
    json={
        "model": "google/gemini-3.1-flash-tts-preview",
        "input": "开桥 is an LLM API aggregation service.",
        "voice": "Kore",
        "response_format": "pcm",
    },
)

# Decode the base64 audio into raw bytes.
with open("speech.pcm", "wb") as f:
    f.write(base64.b64decode(response.json()["audio"]))

{
  "model": "google/gemini-3.1-flash-tts-preview",
  "audio": "<base64-encoded audio bytes>",
  "mime_type": "audio/L16;rate=24000",
  "usage": {
    "input_tokens": 5,
    "output_tokens": 109,
    "total_tokens": 114
  }
}

To receive the audio as Server-Sent Events instead, set stream to true:

curl https://openqiao.com/api/v1/audio/speech \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENQIAO_API_KEY" \
  -d '{
    "model": "google/gemini-3.1-flash-tts-preview",
    "input": "开桥 is an LLM API aggregation service.",
    "voice": "Kore",
    "response_format": "pcm",
    "stream": true
  }'

Content-Type: text/event-stream

data: {"type":"speech.audio.delta","audio":"<base64-encoded audio bytes>","mime_type":"audio/L16;rate=24000"}

data: {"type":"speech.audio.done","usage":{"input_tokens":5,"output_tokens":109,"total_tokens":114}}

data: [DONE]