Create speech

Use this interface to generate speech audio from text. The interface format is compatible with the OpenAI audio/speech protocol.

Request Address

Authorization

Authorization

• Type: HTTP Bearer Auth, value type string
• Required
• Description: Used to verify account information. Go to API Console , click Create Token to obtain.

Authorization: Bearer <YOUR_API_KEY>

Request Headers

Content-Type

• Type: enum<string>
• Required, fixed value application/json
• Description: The media type of the request body. Must be application/json.

Request Body

Overview

model

• Type: string
• Required
• Description: The ID of the TTS model to use.
• Options: Ming-3.0-tts

input

• Type: string
• Required
• Description: The text content to synthesize into speech. Maximum length is 2048 characters. Must not be null or an empty string.

voice

• Type: string
• Optional
• Description: The voice used to generate speech. Supports system voices. To use a custom voice, set the references parameter instead.

references

• Type: object[]
• Optional
• Description: A list of reference audio clips used for stateless voice cloning or voice design, which serves as the custom voice for the generated speech. The current model supports at most 1 reference, each up to 30 seconds and no larger than 10 MB, in WAV format.

references[].audio

• Type: string
• Required
• Description: The Base64-encoded audio data of the reference sample. For best results, use 10–30 seconds of clear speech with as little background noise as possible. Supported format: WAV.

Convert the audio file to a Base64-encoded string before passing it in. The encoded string must not exceed 10 MB. Prefix the Base64 string with data:{MIME_TYPE};base64,$BASE64_AUDIO:

• {MIME_TYPE} — the MIME type of the audio file, used to identify its format. Replace it with the actual MIME value, for example audio/wav.
• $BASE64_AUDIO — the pure Base64-encoded string of the audio file (without any prefix).

references[].text

• Type: string
• Required
• Description: The verbatim transcript of what is said in the audio sample. The accuracy of both the audio and this transcript is critical to voice cloning quality.

instructions

• Type: string
• Optional
• Description: Voice control instructions for the generated speech. Use a voice style description so the speech is narrated with the designed voice.

response_format

• Type: string
• Optional, default pcm for streaming, wav for non-streaming
• Description: The output audio format.
• Options: mp3, wav, pcm, opus, aac

sample_rate

• Type: number
• Optional, default 22050 Hz
• Description: The sample rate of the output audio, in Hz.
• Options: 8000, 16000, 22050, 24000

speed

• Type: number
• Optional, default 1.0
• Description: The speed of the generated speech. Range [0.25, 4.0].

volume

• Type: number
• Optional, default 1.0
• Description: The volume of the generated speech. Range [0.5, 2.0].

pitch

• Type: number
• Optional, default 1.0
• Description: The pitch of the generated speech. Range [0.5, 2.0].

stream_format

• Type: string
• Optional
• Description: Enables streaming audio return, suitable for low-latency scenarios that play audio while it is being generated.
• Options:

The sse mode suits browser clients consuming audio via EventSource, with clear event boundaries and an end signal; the audio mode suits server-side or native clients consuming the audio stream directly, with lower overhead.

Response Body

Non-Streaming

The response is of type file — audio bytes. The Content-Type of the HTTP response varies with the response_format parameter:

SSE Stream

• Content-Type: text/event-stream
• Description: Server-Sent Events stream. Each frame is data: <json>\n\n, where the JSON payload matches a TTS Event.

speech.audio.delta

• Type: object
• Description: An incremental audio chunk event.

data: {"type": "speech.audio.delta", "audio": "//uQxAAAAAANIAAAAAExBTUUzLjEw..."}

• Fields:
  • type — Fixed value speech.audio.delta.
  • audio — A chunk of Base64-encoded audio data.

speech.audio.done

• Type: object
• Description: The terminating event indicating that speech synthesis is complete.

data: {"type": "speech.audio.done", "usage": {"prompt_tokens": 12, "completion_tokens": 80, "total_tokens": 92}}

• Fields:
  • type — Fixed value speech.audio.done.
  • usage — Token usage statistics: prompt_tokens (input), completion_tokens (output), total_tokens (total).

Audio Byte Stream

• Content-Type: Corresponds to response_format (see the table above)
• Description: The response is a raw audio bytestream streamed via HTTP chunked transfer encoding.

Example

Request Example

curl --request POST \
  --url https://api.ant-ling.com/v1/audio/speech \
  --header 'Authorization: Bearer <YOUR_API_KEY>' \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "Ming-3.0-tts",
  "input": "Hello, who are you?",
  "voice": "UK_Charlotte",
  "response_format": "pcm",
  "stream_format": "sse"
}'

from openai import OpenAI
 
client = OpenAI(
    base_url="https://api.ant-ling.com/v1",
    api_key="<YOUR_API_KEY>"
)
 
speech = client.audio.speech.create(
    model="Ming-3.0-tts",
    input="Hello, who are you?",
    voice="UK_Charlotte",
    response_format="pcm",
    extra_body={"stream_format": "sse"}
)
 
speech.stream_to_file("/path/to/output.pcm")

import OpenAI from 'openai';
import fs from 'fs';
 
const client = new OpenAI({
  baseURL: 'https://api.ant-ling.com/v1',
  apiKey: '<YOUR_API_KEY>',
});
 
const response = await client.audio.speech.create({
  model: 'Ming-3.0-tts',
  input: 'Hello, who are you?',
  voice: 'UK_Charlotte',
  response_format: 'pcm',
  stream_format: 'sse',
});
 
response.body.pipe(fs.createWriteStream('/path/to/output.pcm'));

curl --request POST \
  --url https://api.ant-ling.com/v1/audio/speech \
  --header 'Authorization: Bearer <YOUR_API_KEY>' \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "Ming-3.0-tts",
  "input": "Hello, who are you?",
  "references": [
    {
      "audio": "data:audio/wav;base64,$BASE64_AUDIO",
      "text": "Our vision is to build the digital infrastructure of the future service industry and bring more small and beautiful changes to the world."
    }
  ],
  "response_format": "pcm",
  "stream_format": "sse"
}'

Response Example

HTTP/1.1 200 OK
Content-Type: audio/wav

<binary audio data>

data: {"type": "speech.audio.delta", "audio": "//uQxAAAAAANIAAAAAExBTUUzLjEw..."}

data: {"type": "speech.audio.done", "usage": {"prompt_tokens": 12, "completion_tokens": 80, "total_tokens": 92}}

data: {"type": "speech.audio.delta", "audio": "//uQxAAAAAANIAAAAAExBTUUzLjEw..."}

data: {"type": "speech.audio.delta", "audio": "VVVVVVVVVVVVVVVVVVVVVVVVVVVV..."}

...

data: {"type": "speech.audio.done", "usage": {"prompt_tokens": 12, "completion_tokens": 80, "total_tokens": 92}}

HTTP/1.1 200 OK
Content-Type: audio/mpeg
Transfer-Encoding: chunked

<audio binary chunk 1>
<audio binary chunk 2>
...