Build Multimodal Voice & Vision Apps with Qwen-Omni API - Model Studio

The Qwen-Omni model accepts text combined with one other modality, such as an image, audio, or video, and generates responses in text or speech. It offers multiple human-like voices and supports multilingual and dialectal speech output. You can use it for applications such as text creation, visual recognition, and voice assistants.

QuickStart

Prerequisites

Configure an API key and set it as an environment variable.
Qwen-Omni supports only OpenAI-compatible calls. You must install the latest SDK. The minimum required versions are as follows:

Call method: Qwen-Omni supports only streaming output. You must set the stream parameter to True. Non-streaming calls will fail.

The following example sends a text prompt to the Qwen-Omni API and returns a streaming response that contains both text and audio.

Python

import os
import base64
import soundfile as sf
import numpy as np
from openai import OpenAI

# 1. Initialize the client
client = OpenAI(
    api_key=os.getenv("DASHSCOPE_API_KEY"),  # Confirm the environment variable is set
    base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)

# 2. Send the request
try:
    completion = client.chat.completions.create(
        model="qwen3-omni-flash",
        messages=[{"role": "user", "content": "Who are you?"}],
        modalities=["text", "audio"],  # Specify text and audio output
        audio={"voice": "Cherry", "format": "wav"},
        stream=True,  # Must be set to True
        stream_options={"include_usage": True},
    )

    # 3. Process the streaming response and decode the audio
    print("Model response:")
    audio_base64_string = ""
    for chunk in completion:
        # Process the text part
        if chunk.choices and chunk.choices[0].delta.content:
            print(chunk.choices[0].delta.content, end="")

        # Collect the audio part
        if chunk.choices and hasattr(chunk.choices[0].delta, "audio") and chunk.choices[0].delta.audio:
            audio_base64_string += chunk.choices[0].delta.audio.get("data", "")

    # 4. Save the audio file
    if audio_base64_string:
        wav_bytes = base64.b64decode(audio_base64_string)
        audio_np = np.frombuffer(wav_bytes, dtype=np.int16)
        sf.write("audio_assistant.wav", audio_np, samplerate=24000)
        print("\nAudio file saved to: audio_assistant.wav")

except Exception as e:
    print(f"Request failed: {e}")

Node.js

// Setup instructions:
// Universal for Windows/Mac/Linux:
// 1. Ensure Node.js is installed (version >= 14 recommended)
// 2. Run the following command to install required dependencies:
//    npm install openai wav

import OpenAI from "openai";
import { createWriteStream } from 'node:fs';
import { Writer } from 'wav';

// Define an audio conversion function: convert a Base64 string and save it as a standard WAV audio file
async function convertAudio(audioString, audioPath) {
    try {
        // Decode the Base64 string into a Buffer
        const wavBuffer = Buffer.from(audioString, 'base64');
        // Create a WAV file write stream
        const writer = new Writer({
            sampleRate: 24000,  // Sample rate
            channels: 1,        // Mono
            bitDepth: 16        // 16-bit depth
        });
        // Create an output file stream and establish a pipeline connection
        const outputStream = createWriteStream(audioPath);
        writer.pipe(outputStream);

        // Write PCM data and end writing
        writer.write(wavBuffer);
        writer.end();

        // Use a Promise to wait for the file to finish writing
        await new Promise((resolve, reject) => {
            outputStream.on('finish', resolve);
            outputStream.on('error', reject);
        });

        // Add extra wait time to ensure audio integrity
        await new Promise(resolve => setTimeout(resolve, 800));

        console.log(`\nAudio file saved to: ${audioPath}`);
    } catch (error) {
        console.error('Error during audio processing:', error);
    }
}

// 1. Initialize the client
const openai = new OpenAI(
    {
        // The API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
        apiKey: process.env.DASHSCOPE_API_KEY,
        // The following URL is for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
        baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
    }
);
// 2. Send the request
const completion = await openai.chat.completions.create({
    model: "qwen3-omni-flash",  
    messages: [
        {
            "role": "user",
            "content": "Who are you?"
        }],
    stream: true,
    stream_options: {
        include_usage: true
    },
    modalities: ["text", "audio"],
    audio: { voice: "Cherry", format: "wav" }
});

let audioString = "";
console.log("Large language model response:")

// 3. Process the streaming response and decode the audio
for await (const chunk of completion) {
    if (Array.isArray(chunk.choices) && chunk.choices.length > 0) {
        // Process text content
        if (chunk.choices[0].delta.content) {
            process.stdout.write(chunk.choices[0].delta.content);
        }
        // Process audio content
        if (chunk.choices[0].delta.audio) {
            if (chunk.choices[0].delta.audio["data"]) {
                audioString += chunk.choices[0].delta.audio["data"];
            }
        }
    }
}
// 4. Save the audio file
convertAudio(audioString, "audio_assistant.wav");

curl

# ======= Important note =======
# API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
# The following URL is for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions
# === Delete this comment before execution ===

curl -X POST https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions \
-H "Authorization: Bearer $DASHSCOPE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
    "model": "qwen3-omni-flash",
    "messages": [
        {
            "role": "user", 
            "content": "Who are you?"
        }
    ],
    "stream":true,
    "stream_options":{
        "include_usage":true
    },
    "modalities":["text","audio"],
    "audio":{"voice":"Cherry","format":"wav"}
}'

Response

After you run the Python or Node.js code, the model's text response appears in the console, and an audio file named audio_assistant.wav is created in the same directory as your code file.

Large language model response:
I am a large language model developed by Alibaba Cloud. My name is Qwen. How can I help you?

Running the HTTP code directly returns text and Base64-encoded audio data in the audio field.

data: {"choices":[{"delta":{"content":"I"},"finish_reason":null,"index":0,"logprobs":null}],"object":"chat.completion.chunk","usage":null,"created":1757647879,"system_fingerprint":null,"model":"qwen3-omni-flash","id":"chatcmpl-a68eca3b-c67e-4666-a72f-73c0b4919860"}
data: {"choices":[{"delta":{"content":" am"},"finish_reason":null,"index":0,"logprobs":null}],"object":"chat.completion.chunk","usage":null,"created":1757647879,"system_fingerprint":null,"model":"qwen3-omni-flash","id":"chatcmpl-a68eca3b-c67e-4666-a72f-73c0b4919860"}
......
data: {"choices":[{"delta":{"audio":{"data":"/v8AAAAAAAAAAAAAAA...","expires_at":1757647879,"id":"audio_a68eca3b-c67e-4666-a72f-73c0b4919860"}},"finish_reason":null,"index":0,"logprobs":null}],"object":"chat.completion.chunk","usage":null,"created":1757647879,"system_fingerprint":null,"model":"qwen3-omni-flash","id":"chatcmpl-a68eca3b-c67e-4666-a72f-73c0b4919860"}
data: {"choices":[{"finish_reason":"stop","delta":{"content":""},"index":0,"logprobs":null}],"object":"chat.completion.chunk","usage":null,"created":1764763585,"system_fingerprint":null,"model":"qwen3-omni-flash","id":"chatcmpl-e8c82e9e-073e-4289-a786-a20eb444ac9c"}
data: {"choices":[],"object":"chat.completion.chunk","usage":{"prompt_tokens":207,"completion_tokens":103,"total_tokens":310,"completion_tokens_details":{"audio_tokens":83,"text_tokens":20},"prompt_tokens_details":{"text_tokens":207}},"created":1757940330,"system_fingerprint":null,"model":"qwen3-omni-flash","id":"chatcmpl-9cdd5a26-f9e9-4eff-9dcc-93a878165afc"}

Applicable scope

Supported regions

Singapore: Use the API key for this region.
Beijing: Use the API key for this region.

Supported models

Compared to the Qwen-VL models, the Qwen-Omni model can:

Understand visual and audio information in video files.
Understand multimodal data.
Output audio.

It also performs well in visual and audio understanding.

We recommend using qwen3-omni-flash. Compared to qwen-omni-turbo, which is no longer updated, its capabilities have improved significantly:

Supports thinking and non-thinking modes. You can switch between them using the enable_thinking parameter. The thinking mode is disabled by default.
In non-thinking mode, for audio output:
- qwen3-omni-flash-2025-12-01 supports up to 49 voices. qwen3-omni-flash-2025-09-15 and qwen3-omni-flash support up to 17 voices. Qwen-Omni-Turbo supports only 4 voices.
- The number of supported languages has increased to 10. Qwen-Omni-Turbo supports only 2.

International (Singapore)

Commercial models

Commercial models offer newer features and improvements over open-source versions.

Model name	Version	Mode	Context length	Maximum input	Longest Chain of Thought	Maximum output	Free quota (Note)
			(Tokens)
qwen3-omni-flash Capabilities are the same as qwen3-omni-flash-2025-12-01.	Stable version	Thinking mode	65,536	16,384	32,768	16,384	1 million tokens per model, regardless of modality. Valid for 90 days after Model Studio activation.
		Non-thinking mode		49,152	-
qwen3-omni-flash-2025-12-01	Snapshot version	Thinking mode	65,536	16,384	32,768	16,384
		Non-thinking mode		49,152	-
qwen3-omni-flash-2025-09-15 Also known as qwen3-omni-flash-0915.	Snapshot version	Thinking mode	65,536	16,384	32,768	16,384
		Non-thinking mode		49,152	-

More models

Model name	Version	Context length	Maximum input	Maximum output	Free quota (Note)
		(Tokens)
qwen-omni-turbo Matches the capabilities of qwen-omni-turbo-2025-03-26.	Stable version	32,768	30,720	2,048	1 million tokens (not distinguished by modality) Valid for 90 days after Model Studio activation.
qwen-omni-turbo-latest Always matches the latest snapshot version. The capabilities are identical.	Latest version
qwen-omni-turbo-2025-03-26 Also known as qwen-omni-turbo-0326.	Snapshot version

Open source models

Model

Context window

Max input

Max output

Free quota

Note

(tokens)

qwen2.5-omni-7b

32,768

30,720

2,048

1 million tokens (no modality distinction)

Valid for 90 days after activating Model Studio

Mainland China (Beijing)

Commercial models

Model name	Version	Mode	Context length	Maximum input	Maximum chain-of-thought	Maximum output	Free quota (Note)
			(Tokens)
qwen3-omni-flash Currently matches qwen3-omni-flash-2025-12-01 in capability.	Stable version	Thinking mode	65,536	16,384	32,768	16,384	No free quota
		Non-thinking mode		49,152	-
qwen3-omni-flash-2025-12-01	Snapshot version	Thinking mode	65,536	16,384	32,768	16,384
		Non-thinking mode		49,152	-
qwen3-omni-flash-2025-09-15 Also known as qwen3-omni-flash-0915.	Snapshot version	Thinking mode	65,536	16,384	32,768	16,384
		Non-thinking mode		49,152	-

More models

Model name	Version	Context length	Maximum input	Maximum output	Free quota (Note)
		(Tokens)
qwen-omni-turbo Equivalent to qwen-omni-turbo-2025-03-26.	Stable version	32,768	30,720	2,048	No free quota
qwen-omni-turbo-latest Always matches the latest snapshot version. The capabilities are identical.	Latest version
qwen-omni-turbo-2025-03-26 Also known as qwen-omni-turbo-0326.	Snapshot version
qwen-omni-turbo-2025-01-19 Also known as qwen-omni-turbo-0119.

Open source models

Model name

Context window

Max input

Max output

Free quota

(Note)

(tokens)

qwen2.5-omni-7b

32,768

30,720

2,048

No free quota

Usage

Input

Supported input modalities:

In a single user message, the content array can contain text and only one other modality, such as an image, audio, or video. Multiple non-text modalities are not supported.

Methods for inputting multimodal data:
- Public network URL
- Base64 encoding. For more information, see Send local files with Base64 encoding.

Output

Supported output modalities: Audio output is Base64-encoded data. For more information about how to convert it to an audio file, see Work with audio output.

Output modality

modalitiesparameter value

Response style

Text

["text"] (default)

Formal and written style.

Text + audio

["text","audio"]

Qwen3-Omni-Flash does not support audio output in thinking mode.

Conversational style with filler words and interactive prompts.

Qwen-Omni-Turbo does not support setting a system message when audio is included in the output modality.

Supported audio output languages:
- Qwen-Omni-Turbo: Chinese (Mandarin) and English only.
- Qwen3-Omni-Flash (non-thinking mode): Chinese (Mandarin and some dialects), English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, and Korean.
Supported voices: You can configure the voice and file format using the audio parameter. For example, audio={"voice": "Cherry", "format": "wav"}:
- File format (format): This parameter must be set to "wav".
- Voice (voice): For a list of supported voices, see Voice list.

Limitations

Streaming output is mandatory: All requests to Qwen-Omni must have stream=True set.
Only qwen3-omni-flash is a hybrid thinking model. For instructions on how to call this model, see Enable or disable thinking mode. In thinking mode, audio output is not supported.

Enable or disable thinking mode

Qwen3-Omni-Flash is a hybrid thinking model. You can control the thinking mode using the enable_thinking parameter:

true: Enable thinking mode
false (default): Disable thinking mode

Qwen-Omni-Turbo is not a thinking model.

OpenAI compatible

Python

import os
from openai import OpenAI

client = OpenAI(
    # API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    # The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
    base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)

completion = client.chat.completions.create(
    model="qwen3-omni-flash",
    messages=[{"role": "user", "content": "Who are you?"}],

    # Enable or disable thinking mode. Audio output is not supported in thinking mode. qwen-omni-turbo does not support enable_thinking.
    extra_body={'enable_thinking': True},

    # Set the output modality. Currently supported in non-thinking mode: ["text","audio"] and ["text"]. Only ["text"] is supported in thinking mode.
    modalities=["text"],

    # Set the voice. The audio parameter is not supported in thinking mode.
    # audio={"voice": "Cherry", "format": "wav"},
    # stream must be set to True. Otherwise, an error occurs.
    stream=True,
    stream_options={"include_usage": True},
)

for chunk in completion:
    if chunk.choices:
        print(chunk.choices[0].delta)
    else:
        print(chunk.usage)

Node.js

import OpenAI from "openai";

const openai = new OpenAI(
    {
        // API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
        apiKey: process.env.DASHSCOPE_API_KEY,
        // The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
        baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
    }
);
const completion = await openai.chat.completions.create({
    model: "qwen3-omni-flash",
    messages: [
        { role: "user", content: "Who are you?" }
    ],

    // stream must be set to True. Otherwise, an error occurs.
    stream: true,
    stream_options: {
        include_usage: true
    },
    // Enable or disable thinking mode. Audio output is not supported in thinking mode. qwen-omni-turbo does not support enable_thinking.
    extra_body:{'enable_thinking': true},
    // Set the output modality. Currently supported in non-thinking mode: ["text","audio"] and ["text"]. Only ["text"] is supported in thinking mode.
    modalities: ["text"],
    // Set the voice. The audio parameter is not supported in thinking mode.
    //audio: { voice: "Cherry", format: "wav" }
});

for await (const chunk of completion) {
    if (Array.isArray(chunk.choices) && chunk.choices.length > 0) {
        console.log(chunk.choices[0].delta);
    } else {
        console.log(chunk.usage);
    }
}

curl

# ======= Important note =======
# API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
# The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions
# === Delete this comment before execution ===

curl -X POST https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions \
-H "Authorization: Bearer $DASHSCOPE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
    "model": "qwen3-omni-flash",
    "messages": [
        {
            "role": "user", 
            "content": "Who are you?"
        }
    ],
    "stream":true,
    "stream_options":{
        "include_usage":true
    },
    "modalities":["text"],
    "enable_thinking": true
}'

Response

data: {"choices":[{"delta":{"content":null,"role":"assistant","reasoning_content":""},"index":0,"logprobs":null,"finish_reason":null}],"object":"chat.completion.chunk","usage":null,"created":1757937336,"system_fingerprint":null,"model":"qwen3-omni-flash","id":"chatcmpl-ce3d6fe5-e717-4b7e-8b40-3aef12288d4c"}
data: {"choices":[{"finish_reason":null,"logprobs":null,"delta":{"content":null,"reasoning_content":"Hmm"},"index":0}],"object":"chat.completion.chunk","usage":null,"reated":1757937336,"system_fingerprint":null,"model":"qwen3-omni-flash","id":"chatcmpl-ce3d6fe5-e717-4b7e-8b40-3aef12288d4c"}
data: {"choices":[{"delta":{"content":null,"reasoning_content":","},"finish_reason":null,"index":0,"logprobs":null}],"object":"chat.completion.chunk","usage":null,"reated":1757937336,"system_fingerprint":null,"model":"qwen3-omni-flash","id":"chatcmpl-ce3d6fe5-e717-4b7e-8b40-3aef12288d4c"}
......
data: {"choices":[{"delta":{"content":"Tell me"},"finish_reason":null,"index":0,"logprobs":null}],"object":"chat.completion.chunk","usage":null,"created":1757937336,"tem_fingerprint":null,"model":"qwen3-omni-flash","id":"chatcmpl-ce3d6fe5-e717-4b7e-8b40-3aef12288d4c"}
data: {"choices":[{"delta":{"content":"!"},"finish_reason":null,"index":0,"logprobs":null}],"object":"chat.completion.chunk","usage":null,"created":1757937336,"systm_fingerprint":null,"model":"qwen3-omni-flash","id":"chatcmpl-ce3d6fe5-e717-4b7e-8b40-3aef12288d4c"}
data: {"choices":[{"finish_reason":"stop","delta":{"content":"","reasoning_content":null},"index":0,"logprobs":null}],"object":"chat.completion.chunk","usage":null,"created":1757937336,"system_fingerprint":null,"model":"qwen3-omni-flash","id":"chatcmpl-ce3d6fe5-e717-4b7e-8b40-3aef12288d4c"}
data: {"choices":[],"object":"chat.completion.chunk","usage":{"prompt_tokens":11,"completion_tokens":363,"total_tokens":374,"completion_tokens_details":{"reasoning_tokens":195,"text_tokens":168},"prompt_tokens_details":{"text_tokens":11}},"created":1757937336,"system_fingerprint":null,"model":"qwen3-omni-flash","id":"chatcmpl-ce3d6fe5-e717-4b7e-8b40-3aef12288d4c"}

Image + text input

Qwen-Omni supports multiple images per request. The image requirements are as follows:

Each image file must be no larger than 10 MB.
Number of images: Up to 2048 images when using public network URLs or local paths. Up to 250 images when using Base64 encoding.
The total number of tokens for images and text must not exceed the model's maximum input limit.
Both the width and height must exceed 10 pixels. The aspect ratio must not exceed 200:1 or 1:200.
For a list of supported image types, see Visual and video understanding.

The following examples use a public network image URL. To use a local image instead, see Send local files with Base64 encoding. Streaming output is required for all calls.

OpenAI compatible

Python

import os
from openai import OpenAI

client = OpenAI(
    # API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    # The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
    base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)

completion = client.chat.completions.create(
    model="qwen3-omni-flash", # When using qwen3-omni-flash, run in non-thinking mode.
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "type": "image_url",
                    "image_url": {
                        "url": "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241022/emyrja/dog_and_girl.jpeg"
                    },
                },
                {"type": "text", "text": "What scene is depicted in the image?"},
            ],
        },
    ],
    # Set the output modality. Currently supported: ["text","audio"] and ["text"].
    modalities=["text", "audio"],
    audio={"voice": "Cherry", "format": "wav"},
    # stream must be set to True. Otherwise, an error occurs.
    stream=True,
    stream_options={
        "include_usage": True
    }
)

for chunk in completion:
    if chunk.choices:
        print(chunk.choices[0].delta)
    else:
        print(chunk.usage)

Node.js

import OpenAI from "openai";

const openai = new OpenAI(
    {
        // API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
        apiKey: process.env.DASHSCOPE_API_KEY,
        // The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
        baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
    }
);
const completion = await openai.chat.completions.create({
    model: "qwen3-omni-flash", // When using qwen3-omni-flash, run in non-thinking mode.
    messages: [
        {
            "role": "user",
            "content": [{
                "type": "image_url",
                "image_url": { "url": "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241022/emyrja/dog_and_girl.jpeg" },
            },
            { "type": "text", "text": "What scene is depicted in the image?" }]
        }],
    stream: true,
    stream_options: {
        include_usage: true
    },
    modalities: ["text", "audio"],
    audio: { voice: "Cherry", format: "wav" }
});

for await (const chunk of completion) {
    if (Array.isArray(chunk.choices) && chunk.choices.length > 0) {
        console.log(chunk.choices[0].delta);
    } else {
        console.log(chunk.usage);
    }
}

curl

# ======= Important note =======
# API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
# The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions
# === Delete this comment before execution ===


curl -X POST https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions \
-H "Authorization: Bearer $DASHSCOPE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
    "model": "qwen3-omni-flash",
    "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "image_url",
          "image_url": {
            "url": "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241022/emyrja/dog_and_girl.jpeg"
          }
        },
        {
          "type": "text",
          "text": "What scene is depicted in the image?"
        }
      ]
    }
  ],
    "stream":true,
    "stream_options":{
        "include_usage":true
    },
    "modalities":["text","audio"],
    "audio":{"voice":"Cherry","format":"wav"}
}'

Audio + text input

You can send only one audio file per request.
File size:
- qwen3-omni-flash: Up to 100 MB. Maximum duration: 20 minutes.
- qwen-omni-turbo: Up to 10 MB. Maximum duration: 3 minutes.
Supported formats: AMR, WAV, 3GP, 3GPP, AAC, and MP3.

The following examples use a public network audio URL. To use a local audio file instead, see Send local files with Base64 encoding. Streaming output is required for all calls.

OpenAI compatible

Python

import os
from openai import OpenAI

client = OpenAI(
    # API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    # The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
    base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)

completion = client.chat.completions.create(
    model="qwen3-omni-flash",# When using qwen3-omni-flash, run in non-thinking mode.
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "type": "input_audio",
                    "input_audio": {
                        "data": "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20250211/tixcef/cherry.wav",
                        "format": "wav",
                    },
                },
                {"type": "text", "text": "What is this audio about?"},
            ],
        },
    ],
    # Set the output modality. Currently supported: ["text","audio"] and ["text"].
    modalities=["text", "audio"],
    audio={"voice": "Cherry", "format": "wav"},
    # stream must be set to True. Otherwise, an error occurs.
    stream=True,
    stream_options={"include_usage": True},
)

for chunk in completion:
    if chunk.choices:
        print(chunk.choices[0].delta)
    else:
        print(chunk.usage)

Node.js

import OpenAI from "openai";

const openai = new OpenAI(
    {
        // API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
        apiKey: process.env.DASHSCOPE_API_KEY,
        // The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
        baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
    }
);
const completion = await openai.chat.completions.create({
    model: "qwen3-omni-flash", // When using qwen3-omni-flash, run in non-thinking mode.
    messages: [
        {
            "role": "user",
            "content": [{
                "type": "input_audio",
                "input_audio": { "data": "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20250211/tixcef/cherry.wav", "format": "wav" },
            },
            { "type": "text", "text": "What is this audio about?" }]
        }],
    stream: true,
    stream_options: {
        include_usage: true
    },
    modalities: ["text", "audio"],
    audio: { voice: "Cherry", format: "wav" }
});

for await (const chunk of completion) {
    if (Array.isArray(chunk.choices) && chunk.choices.length > 0) {
        console.log(chunk.choices[0].delta);
    } else {
        console.log(chunk.usage);
    }
}

curl

# ======= Important note =======
# API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
# The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions
# === Delete this comment before execution ===

curl -X POST https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions \
-H "Authorization: Bearer $DASHSCOPE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
    "model": "qwen3-omni-flash",
    "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "input_audio",
          "input_audio": {
            "data": "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20250211/tixcef/cherry.wav",
            "format": "wav"
          }
        },
        {
          "type": "text",
          "text": "What is this audio about?"
        }
      ]
    }
  ],
    "stream":true,
    "stream_options":{
        "include_usage":true
    },
    "modalities":["text","audio"],
    "audio":{"voice":"Cherry","format":"wav"}
}'

Video + text input

The video input method can be an image list format or a video file format, which can interpret audio in the video.

The following examples use a public network video URL. To use a local video instead, see Send local files with Base64 encoding. Streaming output is required for all calls.

Image list format

Number of images

qwen3-omni-flash: Minimum: 2 images. Maximum: 128 images.
qwen-omni-turbo: Minimum: 4 images. Maximum: 80 images.

OpenAI compatible

Python

import os
from openai import OpenAI

client = OpenAI(
    # API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    # The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
    base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)

completion = client.chat.completions.create(
    model="qwen3-omni-flash", # When using qwen3-omni-flash, run in non-thinking mode.
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "type": "video",
                    "video": [
                        "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/xzsgiz/football1.jpg",
                        "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/tdescd/football2.jpg",
                        "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/zefdja/football3.jpg",
                        "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/aedbqh/football4.jpg",
                    ],
                },
                {"type": "text", "text": "Describe the process shown in this video."},
            ],
        }
    ],
    # Set the output modality. Currently supported: ["text","audio"] and ["text"].
    modalities=["text", "audio"],
    audio={"voice": "Cherry", "format": "wav"},
    # stream must be set to True. Otherwise, an error occurs.
    stream=True,
    stream_options={"include_usage": True},
)

for chunk in completion:
    if chunk.choices:
        print(chunk.choices[0].delta)
    else:
        print(chunk.usage)

Node.js

import OpenAI from "openai";

const openai = new OpenAI(
    {
        // API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
        apiKey: process.env.DASHSCOPE_API_KEY,
        // The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
        baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
    }
);
const completion = await openai.chat.completions.create({
    model: "qwen3-omni-flash", // When using qwen3-omni-flash, run in non-thinking mode.
    messages: [{
        role: "user",
        content: [
            {
                type: "video",
                video: [
                    "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/xzsgiz/football1.jpg",
                    "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/tdescd/football2.jpg",
                    "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/zefdja/football3.jpg",
                    "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/aedbqh/football4.jpg"
                ]
            },
            {
                type: "text",
                text: "Describe the process shown in this video."
            }
        ]
    }],
    stream: true,
    stream_options: {
        include_usage: true
    },
    modalities: ["text", "audio"],
    audio: { voice: "Cherry", format: "wav" }
});

for await (const chunk of completion) {
    if (Array.isArray(chunk.choices) && chunk.choices.length > 0) {
        console.log(chunk.choices[0].delta);
    } else {
        console.log(chunk.usage);
    }
}

curl

# ======= Important note =======
# API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
# The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions
# === Delete this comment before execution ===

curl -X POST https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions \
-H "Authorization: Bearer $DASHSCOPE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
    "model": "qwen3-omni-flash",
    "messages": [
        {
            "role": "user",
            "content": [
                {
                    "type": "video",
                    "video": [
                        "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/xzsgiz/football1.jpg",
                        "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/tdescd/football2.jpg",
                        "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/zefdja/football3.jpg",
                        "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241108/aedbqh/football4.jpg"
                    ]
                },
                {
                    "type": "text",
                    "text": "Describe the process shown in this video."
                }
            ]
        }
    ],
    "stream": true,
    "stream_options": {
        "include_usage": true
    },
    "modalities": ["text", "audio"],
    "audio": {
        "voice": "Cherry",
        "format": "wav"
    }
}'

Video file format (processes audio in the video)

You can send only one video file per request.
File size:
- qwen3-omni-flash: Max 256 MB. Max duration: 150 seconds.
- qwen-omni-turbo: Max 150 MB. Max duration: 40 seconds.
Supported formats: MP4, AVI, MKV, MOV, FLV, WMV, and more.
Visual and audio information in the video file are billed separately.

OpenAI compatible

Python

import os
from openai import OpenAI

client = OpenAI(
    # API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    # The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
    base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)

completion = client.chat.completions.create(
    model="qwen3-omni-flash", # When using qwen3-omni-flash, run in non-thinking mode.
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "type": "video_url",
                    "video_url": {
                        "url": "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241115/cqqkru/1.mp4"
                    },
                },
                {"type": "text", "text": "What is the content of the video?"},
            ],
        },
    ],
    # Set the output modality. Currently supported: ["text","audio"] and ["text"].
    modalities=["text", "audio"],
    audio={"voice": "Cherry", "format": "wav"},
    # stream must be set to True. Otherwise, an error occurs.
    stream=True,
    stream_options={"include_usage": True},
)

for chunk in completion:
    if chunk.choices:
        print(chunk.choices[0].delta)
    else:
        print(chunk.usage)

Node.js

import OpenAI from "openai";

const openai = new OpenAI(
    {
        // API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
        apiKey: process.env.DASHSCOPE_API_KEY,
        // The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
        baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
    }
);
const completion = await openai.chat.completions.create({
    model: "qwen3-omni-flash", // When using qwen3-omni-flash, run in non-thinking mode.
    messages: [
        {
            "role": "user",
            "content": [{
                "type": "video_url",
                "video_url": { "url": "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241115/cqqkru/1.mp4" },
            },
            { "type": "text", "text": "What is the content of the video?" }]
        }],
    stream: true,
    stream_options: {
        include_usage: true
    },
    modalities: ["text", "audio"],
    audio: { voice: "Cherry", format: "wav" }
});


for await (const chunk of completion) {
    if (Array.isArray(chunk.choices) && chunk.choices.length > 0) {
        console.log(chunk.choices[0].delta);
    } else {
        console.log(chunk.usage);
    }
}

curl

# ======= Important note =======
# API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
# The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions
# === Delete this comment before execution ===

curl -X POST https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions \
-H "Authorization: Bearer $DASHSCOPE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
    "model": "qwen3-omni-flash",
    "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "video_url",
          "video_url": {
            "url": "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241115/cqqkru/1.mp4"
          }
        },
        {
          "type": "text",
          "text": "What is the content of the video?"
        }
      ]
    }
  ],
    "stream":true,
    "stream_options": {
        "include_usage": true
    },
    "modalities":["text","audio"],
    "audio":{"voice":"Cherry","format":"wav"}
}'

Multi-turn conversation

When you use Qwen-Omni for multi-turn conversations, you must follow these rules:

Assistant Message
An Assistant Message, added to the messages array, can contain only text data.
User message
A user message can contain only text or data of a single modality. In a multi-turn conversation, you can use different modalities across different user messages.

OpenAI compatible

Python

import os
from openai import OpenAI

client = OpenAI(
    # API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    # The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
    base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)

completion = client.chat.completions.create(
    model="qwen3-omni-flash", # When using qwen3-omni-flash, run in non-thinking mode.
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "type": "input_audio",
                    "input_audio": {
                        "data": "https://dashscope.oss-cn-beijing.aliyuncs.com/audios/welcome.mp3",
                        "format": "mp3",
                    },
                },
                {"type": "text", "text": "What is this audio about?"},
            ],
        },
        {
            "role": "assistant",
            "content": [{"type": "text", "text": "This audio says: Welcome to Alibaba Cloud"}],
        },
        {
            "role": "user",
            "content": [{"type": "text", "text": "Can you tell me about this company?"}],
        },
    ],
    # Set the output modality. Currently supported: ["text","audio"] and ["text"].
    modalities=["text"],
    # stream must be set to True. Otherwise, an error occurs.
    stream=True,
    stream_options={"include_usage": True},
)

for chunk in completion:
    if chunk.choices:
        print(chunk.choices[0].delta)
    else:
        print(chunk.usage)

Node.js

import OpenAI from "openai";

const openai = new OpenAI(
    {
        // API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
        apiKey: process.env.DASHSCOPE_API_KEY,
        // The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
        baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
    }
);
const completion = await openai.chat.completions.create({
    model: "qwen3-omni-flash", // When using qwen3-omni-flash, run in non-thinking mode.
    messages: [
        {
            "role": "user",
            "content": [
                {
                    "type": "input_audio",
                    "input_audio": {
                        "data": "https://dashscope.oss-cn-beijing.aliyuncs.com/audios/welcome.mp3",
                        "format": "mp3",
                    },
                },
                { "type": "text", "text": "What is this audio about?" }
            ],
        },
        {
            "role": "assistant",
            "content": [{ "type": "text", "text": "This audio says: Welcome to Alibaba Cloud" }],
        },
        {
            "role": "user",
            "content": [{ "type": "text", "text": "Can you tell me about this company?" }]
        }],
    stream: true,
    stream_options: {
        include_usage: true
    },
    modalities: ["text"]
});


for await (const chunk of completion) {
    if (Array.isArray(chunk.choices) && chunk.choices.length > 0) {
        console.log(chunk.choices[0].delta);
    } else {
        console.log(chunk.usage);
    }
}

curl

# ======= Important note =======
# API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
# The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions
# === Delete this comment before execution ===

curl -X POST https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions \
-H "Authorization: Bearer $DASHSCOPE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
  "model": "qwen3-omni-flash",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "input_audio",
          "input_audio": {
            "data": "https://dashscope.oss-cn-beijing.aliyuncs.com/audios/welcome.mp3"
          }
        },
        {
          "type": "text",
          "text": "What is this audio about?"
        }
      ]
    },
    {
      "role": "assistant",
      "content": [
        {
          "type": "text",
          "text": "This audio says: Welcome to Alibaba Cloud"
        }
      ]
    },
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "Can you tell me about this company?"
        }
      ]
    }
  ],
  "stream": true,
  "stream_options": {
    "include_usage": true
  },
  "modalities": ["text"]
}'

Parsing Base64-encoded Audio Data Output

Method 1: Decode after generation completes

The Qwen-Omni model outputs audio as Base64-encoded streaming output. You can maintain a string variable during model generation and append the Base64-encoded data from each returned chunk to it. After generation completes, Base64-decode the full string to obtain the audio file. Alternatively, you can decode and play the Base64-encoded data from each returned chunk in real time.

Python

# Installation instructions for pyaudio:
# APPLE Mac OS X
#   brew install portaudio
#   pip install pyaudio
# Debian/Ubuntu
#   sudo apt-get install python-pyaudio python3-pyaudio
#   or
#   pip install pyaudio
# CentOS
#   sudo yum install -y portaudio portaudio-devel && pip install pyaudio
# Microsoft Windows
#   python -m pip install pyaudio

import os
from openai import OpenAI
import base64
import numpy as np
import soundfile as sf

client = OpenAI(
    # API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    # The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
    base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)

completion = client.chat.completions.create(
    model="qwen3-omni-flash", # When using qwen3-omni-flash, run in non-thinking mode.
    messages=[{"role": "user", "content": "Who are you?"}],
    # Set the output modality. Currently supported: ["text","audio"] and ["text"].
    modalities=["text", "audio"],
    audio={"voice": "Cherry", "format": "wav"},
    # stream must be set to True. Otherwise, an error occurs.
    stream=True,
    stream_options={"include_usage": True},
)

# Method 1: Decode after generation completes
audio_string = ""
for chunk in completion:
    if chunk.choices:
        if hasattr(chunk.choices[0].delta, "audio"):
            try:
                audio_string += chunk.choices[0].delta.audio["data"]
            except Exception as e:
                print(chunk.choices[0].delta.content)
    else:
        print(chunk.usage)

wav_bytes = base64.b64decode(audio_string)
audio_np = np.frombuffer(wav_bytes, dtype=np.int16)
sf.write("audio_assistant_py.wav", audio_np, samplerate=24000)

# Method 2: Decode while generating (comment out Method 1 code to use Method 2)
# # Initialize PyAudio
# import pyaudio
# import time
# p = pyaudio.PyAudio()
# # Create an audio stream
# stream = p.open(format=pyaudio.paInt16,
#                 channels=1,
#                 rate=24000,
#                 output=True)

# for chunk in completion:
#     if chunk.choices:
#         if hasattr(chunk.choices[0].delta, "audio"):
#             try:
#                 audio_string = chunk.choices[0].delta.audio["data"]
#                 wav_bytes = base64.b64decode(audio_string)
#                 audio_np = np.frombuffer(wav_bytes, dtype=np.int16)
#                 # Play audio data directly
#                 stream.write(audio_np.tobytes())
#             except Exception as e:
#                 print(chunk.choices[0].delta.content)

# time.sleep(0.8)
# # Clean up resources
# stream.stop_stream()
# stream.close()
# p.terminate()

Node.js

// Setup instructions:
// Universal for Windows/Mac/Linux:
// 1. Ensure Node.js is installed (version >= 14 recommended)
// 2. Run the following command to install required dependencies:
//    npm install openai wav
// 
// To use real-time playback (Method 2), also install:
// Windows:
//    npm install speaker
// Mac:
//    brew install portaudio
//    npm install speaker
// Linux (Ubuntu/Debian):
//    sudo apt-get install libasound2-dev
//    npm install speaker

import OpenAI from "openai";

const openai = new OpenAI(
    {
        // API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
        apiKey: process.env.DASHSCOPE_API_KEY,
        // The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
        baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
    }
);
const completion = await openai.chat.completions.create({
    model: "qwen3-omni-flash", // When using qwen3-omni-flash, run in non-thinking mode.
    messages: [
        {
            "role": "user",
            "content": "Who are you?"
        }],
    stream: true,
    stream_options: {
        include_usage: true
    },
    modalities: ["text", "audio"],
    audio: { voice: "Cherry", format: "wav" }
});

// Method 1: Decode after generation completes
// Requires installation: npm install wav
import { createWriteStream } from 'node:fs';  // node:fs is a built-in Node.js module, no installation required
import { Writer } from 'wav';

async function convertAudio(audioString, audioPath) {
    try {
        // Decode the Base64 string into a Buffer
        const wavBuffer = Buffer.from(audioString, 'base64');
        // Create a WAV file write stream
        const writer = new Writer({
            sampleRate: 24000,  // Sample rate
            channels: 1,        // Mono
            bitDepth: 16        // 16-bit depth
        });
        // Create an output file stream and establish a pipeline connection
        const outputStream = createWriteStream(audioPath);
        writer.pipe(outputStream);

        // Write PCM data and end writing
        writer.write(wavBuffer);
        writer.end();

        // Use a Promise to wait for the file to finish writing
        await new Promise((resolve, reject) => {
            outputStream.on('finish', resolve);
            outputStream.on('error', reject);
        });

        // Add extra wait time to ensure audio integrity
        await new Promise(resolve => setTimeout(resolve, 800));

        console.log(`Audio file saved to: ${audioPath}`);
    } catch (error) {
        console.error('Error during audio processing:', error);
    }
}

let audioString = "";
for await (const chunk of completion) {
    if (Array.isArray(chunk.choices) && chunk.choices.length > 0) {
        if (chunk.choices[0].delta.audio) {
            if (chunk.choices[0].delta.audio["data"]) {
                audioString += chunk.choices[0].delta.audio["data"];
            }
        }
    } else {
        console.log(chunk.usage);
    }
}
// Execute conversion
convertAudio(audioString, "audio_assistant_mjs.wav");


// Method 2: Generate and play in real time
// Install required components per system instructions above first.
// import Speaker from 'speaker'; // Import audio playback library

// // Create a speaker instance (configuration matches WAV file parameters)
// const speaker = new Speaker({
//     sampleRate: 24000,  // Sample rate
//     channels: 1,        // Number of sound channels
//     bitDepth: 16,       // Bit depth
//     signed: true        // Signed PCM
// });
// for await (const chunk of completion) {
//     if (Array.isArray(chunk.choices) && chunk.choices.length > 0) {
//         if (chunk.choices[0].delta.audio) {
//             if (chunk.choices[0].delta.audio["data"]) {
//                 const pcmBuffer = Buffer.from(chunk.choices[0].delta.audio.data, 'base64');
//                 // Write directly to speaker for playback
//                 speaker.write(pcmBuffer);
//             }
//         }
//     } else {
//         console.log(chunk.usage);
//     }
// }
// speaker.on('finish', () => console.log('Playback complete'));
// speaker.end(); // Call based on actual API stream end

Input a Base64-encoded local file

Images

This example uses the locally saved file eagle.png.

Python

import os
from openai import OpenAI
import base64

client = OpenAI(
    # API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    # The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
    base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)


# Base64 encoding format
def encode_image(image_path):
    with open(image_path, "rb") as image_file:
        return base64.b64encode(image_file.read()).decode("utf-8")


base64_image = encode_image("eagle.png")

completion = client.chat.completions.create(
    model="qwen3-omni-flash", # When using qwen3-omni-flash, run in non-thinking mode.
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "type": "image_url",
                    "image_url": {"url": f"data:image/png;base64,{base64_image}"},
                },
                {"type": "text", "text": "What scene is depicted in the image?"},
            ],
        },
    ],
    # Set the output modality. Currently supported: ["text","audio"] and ["text"].
    modalities=["text", "audio"],
    audio={"voice": "Cherry", "format": "wav"},
    # stream must be set to True. Otherwise, an error occurs.
    stream=True,
    stream_options={"include_usage": True},
)

for chunk in completion:
    if chunk.choices:
        print(chunk.choices[0].delta)
    else:
        print(chunk.usage)

Node.js

import OpenAI from "openai";
import { readFileSync } from 'fs';

const openai = new OpenAI(
    {
        // API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
        apiKey: process.env.DASHSCOPE_API_KEY,
        // The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
        baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
    }
);

const encodeImage = (imagePath) => {
    const imageFile = readFileSync(imagePath);
    return imageFile.toString('base64');
};
const base64Image = encodeImage("eagle.png")

const completion = await openai.chat.completions.create({
    model: "qwen3-omni-flash",// When using qwen3-omni-flash, run in non-thinking mode.
    messages: [
        {
            "role": "user",
            "content": [{
                "type": "image_url",
                "image_url": { "url": `data:image/png;base64,${base64Image}` },
            },
            { "type": "text", "text": "What scene is depicted in the image?" }]
        }],
    stream: true,
    stream_options: {
        include_usage: true
    },
    modalities: ["text", "audio"],
    audio: { voice: "Cherry", format: "wav" }
});

for await (const chunk of completion) {
    if (Array.isArray(chunk.choices) && chunk.choices.length > 0) {
        console.log(chunk.choices[0].delta);
    } else {
        console.log(chunk.usage);
    }
}

Audio

This example uses the locally saved file welcome.mp3.

Python

import os
from openai import OpenAI
import base64
import numpy as np
import soundfile as sf
import requests

client = OpenAI(
    # API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    # The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
    base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)


def encode_audio(audio_path):
    with open(audio_path, "rb") as audio_file:
        return base64.b64encode(audio_file.read()).decode("utf-8")


base64_audio = encode_audio("welcome.mp3")

completion = client.chat.completions.create(
    model="qwen3-omni-flash", # When using qwen3-omni-flash, run in non-thinking mode.
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "type": "input_audio",
                    "input_audio": {
                        "data": f"data:;base64,{base64_audio}",
                        "format": "mp3",
                    },
                },
                {"type": "text", "text": "What is this audio about?"},
            ],
        },
    ],
    # Set the output modality. Currently supported: ["text","audio"] and ["text"].
    modalities=["text", "audio"],
    audio={"voice": "Cherry", "format": "wav"},
    # stream must be set to True. Otherwise, an error occurs.
    stream=True,
    stream_options={"include_usage": True},
)

for chunk in completion:
    if chunk.choices:
        print(chunk.choices[0].delta)
    else:
        print(chunk.usage)

Node.js

import OpenAI from "openai";
import { readFileSync } from 'fs';

const openai = new OpenAI(
    {
        // API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
        apiKey: process.env.DASHSCOPE_API_KEY,
        // The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
        baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
    }
);

const encodeAudio = (audioPath) => {
    const audioFile = readFileSync(audioPath);
    return audioFile.toString('base64');
};
const base64Audio = encodeAudio("welcome.mp3")

const completion = await openai.chat.completions.create({
    model: "qwen3-omni-flash", // When using qwen3-omni-flash, run in non-thinking mode.
    messages: [
        {
            "role": "user",
            "content": [{
                "type": "input_audio",
                "input_audio": { "data": `data:;base64,${base64Audio}`, "format": "mp3" },
            },
            { "type": "text", "text": "What is this audio about?" }]
        }],
    stream: true,
    stream_options: {
        include_usage: true
    },
    modalities: ["text", "audio"],
    audio: { voice: "Cherry", format: "wav" }
});

for await (const chunk of completion) {
    if (Array.isArray(chunk.choices) && chunk.choices.length > 0) {
        console.log(chunk.choices[0].delta);
    } else {
        console.log(chunk.usage);
    }
}

Video

Video file

This example uses the local file spring_mountain.mp4.

Python

import os
from openai import OpenAI
import base64
import numpy as np
import soundfile as sf

client = OpenAI(
    # API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    # The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
    base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)

# Base64 encoding format
def encode_video(video_path):
    with open(video_path, "rb") as video_file:
        return base64.b64encode(video_file.read()).decode("utf-8")


base64_video = encode_video("spring_mountain.mp4")

completion = client.chat.completions.create(
    model="qwen3-omni-flash", # When using qwen3-omni-flash, run in non-thinking mode.
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "type": "video_url",
                    "video_url": {"url": f"data:;base64,{base64_video}"},
                },
                {"type": "text", "text": "What is she singing?"},
            ],
        },
    ],
    # Set the output modality. Currently supported: ["text","audio"] and ["text"].
    modalities=["text", "audio"],
    audio={"voice": "Cherry", "format": "wav"},
    # stream must be set to True. Otherwise, an error occurs.
    stream=True,
    stream_options={"include_usage": True},
)

for chunk in completion:
    if chunk.choices:
        print(chunk.choices[0].delta)
    else:
        print(chunk.usage)

Node.js

import OpenAI from "openai";
import { readFileSync } from 'fs';

const openai = new OpenAI(
    {
        // API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
        apiKey: process.env.DASHSCOPE_API_KEY,
        // The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
        baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
    }
);

const encodeVideo = (videoPath) => {
    const videoFile = readFileSync(videoPath);
    return videoFile.toString('base64');
};
const base64Video = encodeVideo("spring_mountain.mp4")

const completion = await openai.chat.completions.create({
    model: "qwen3-omni-flash", // When using qwen3-omni-flash, run in non-thinking mode.
    messages: [
        {
            "role": "user",
            "content": [{
                "type": "video_url",
                "video_url": { "url": `data:;base64,${base64Video}` },
            },
            { "type": "text", "text": "What is she singing?" }]
        }],
    stream: true,
    stream_options: {
        include_usage: true
    },
    modalities: ["text", "audio"],
    audio: { voice: "Cherry", format: "wav" }
});

for await (const chunk of completion) {
    if (Array.isArray(chunk.choices) && chunk.choices.length > 0) {
        console.log(chunk.choices[0].delta);
    } else {
        console.log(chunk.usage);
    }
}

Image list

Take the locally stored files football1.jpg, football2.jpg, football3.jpg, and football4.jpg as an example.

Python

import os
from openai import OpenAI
import base64
import numpy as np
import soundfile as sf

client = OpenAI(
    # API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    # The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
    base_url="https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
)


# Base64 encoding format
def encode_image(image_path):
    with open(image_path, "rb") as image_file:
        return base64.b64encode(image_file.read()).decode("utf-8")


base64_image_1 = encode_image("football1.jpg")
base64_image_2 = encode_image("football2.jpg")
base64_image_3 = encode_image("football3.jpg")
base64_image_4 = encode_image("football4.jpg")

completion = client.chat.completions.create(
    model="qwen3-omni-flash", # When using qwen3-omni-flash, run in non-thinking mode.
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "type": "video",
                    "video": [
                        f"data:image/jpeg;base64,{base64_image_1}",
                        f"data:image/jpeg;base64,{base64_image_2}",
                        f"data:image/jpeg;base64,{base64_image_3}",
                        f"data:image/jpeg;base64,{base64_image_4}",
                    ],
                },
                {"type": "text", "text": "Describe the procedure in this video."},
            ],
        }
    ],
    # Set the output modality. Currently supported: ["text","audio"] and ["text"].
    modalities=["text", "audio"],
    audio={"voice": "Cherry", "format": "wav"},
    # stream must be set to True. Otherwise, an error occurs.
    stream=True,
    stream_options={"include_usage": True},
)

for chunk in completion:
    if chunk.choices:
        print(chunk.choices[0].delta)
    else:
        print(chunk.usage)

Node.js

import OpenAI from "openai";
import { readFileSync } from 'fs';

const openai = new OpenAI(
    {
        // API keys for Singapore and Beijing regions differ. To get an API key, see https://www.alibabacloud.com/help/zh/model-studio/get-api-key
        apiKey: process.env.DASHSCOPE_API_KEY,
        // The following is the URL for the Singapore region. If using a Beijing-region model, replace it with: https://dashscope.aliyuncs.com/compatible-mode/v1
        baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
    }
);

const encodeImage = (imagePath) => {
    const imageFile = readFileSync(imagePath);
    return imageFile.toString('base64');
  };
const base64Image1 = encodeImage("football1.jpg")
const base64Image2 = encodeImage("football2.jpg")
const base64Image3 = encodeImage("football3.jpg")
const base64Image4 = encodeImage("football4.jpg")

const completion = await openai.chat.completions.create({
    model: "qwen3-omni-flash", // When using qwen3-omni-flash, run in non-thinking mode.
    messages: [{
        role: "user",
        content: [
            {
                type: "video",
                video: [
                    `data:image/jpeg;base64,${base64Image1}`,
                    `data:image/jpeg;base64,${base64Image2}`,
                    `data:image/jpeg;base64,${base64Image3}`,
                    `data:image/jpeg;base64,${base64Image4}`
                ]
            },
            {
                type: "text",
                text: "Describe the procedure in this video."
            }
        ]
    }],
    stream: true,
    stream_options: {
        include_usage: true
    },
    modalities: ["text", "audio"],
    audio: { voice: "Cherry", format: "wav" }
});

for await (const chunk of completion) {
    if (Array.isArray(chunk.choices) && chunk.choices.length > 0) {
        console.log(chunk.choices[0].delta);
    } else {
        console.log(chunk.usage);
    }

API reference

For the complete specifications of the input and output parameters for the Qwen-Omni model, see Qwen.

Billing and rate limits

Billing rules

Qwen-Omni billing is based on the tokens consumed across different modalities, such as audio, image, and video. For more information about pricing, see Models.

Token conversion rules for audio, images, and video

Audio

qwen3-omni-flash: Total tokens = Audio duration (seconds) × 12.5
qwen-omni-turbo: Total tokens = Audio duration (seconds) × 25. Audio shorter than 1 second is billed as 1 second.

Images

qwen3-omni-flash models: One token per 32×32 pixels.
qwen-omni-turbo models: One token per 28×28 pixels.

A single image requires a minimum of 4 tokens and a maximum of 1280 tokens. You can use the following code to estimate the token count for a given image:

import math
# Install Pillow: pip install Pillow
from PIL import Image

# For qwen-omni-turbo, factor is 28.
# factor = 28
# For qwen3-omni-flash, factor is 32.
factor = 32

def token_calculate(image_path=''):
    """
    param image_path: Path to the image.
    return: Token count for a single image.
    """
    if len(image_path) > 0:
        # Open the specified PNG image file.
        image = Image.open(image_path)
        # Get original dimensions.
        height = image.height
        width = image.width
        print(f"Original image dimensions: Height={height}, Width={width}")
        # Adjust height to multiple of factor.
        h_bar = round(height / factor) * factor
        # Adjust width to multiple of factor.
        w_bar = round(width / factor) * factor
        # Minimum tokens for an image: 4 tokens.
        min_pixels = 4 * factor * factor
        # Maximum tokens for an image: 1280 tokens.
        max_pixels = 1280 * factor * factor
        # Scale image to fit within [min_pixels, max_pixels] range.
        if h_bar * w_bar > max_pixels:
            # Calculate scaling factor beta so total pixels do not exceed max_pixels.
            beta = math.sqrt((height * width) / max_pixels)
            # Recalculate adjusted height to ensure multiple of factor.
            h_bar =math.floor(height / beta / factor) * factor
            # Recalculate adjusted width to ensure it is a multiple of the factor.
            w_bar = math.floor(width / beta / factor) * factor
        elif h_bar * w_bar < min_pixels:
            # Calculate scaling factor beta so that the total pixels of the scaled image are not less than min_pixels.
            beta = math.sqrt(min_pixels / (height * width))
            # Recalculate adjusted height to ensure it is a multiple of the factor.
            h_bar = math.ceil(height * beta / factor) * factor
            # Recalculate adjusted width to ensure it is a multiple of the factor.
            w_bar = math.ceil(width * beta / factor) * factor
        print(f"Image dimensions after scaling: Height={h_bar}, Width={w_bar}")
        # Calculate image tokens: total pixels / (factor * factor).
        token = int((h_bar * w_bar) / (factor * factor)) + 2
        print(f"Token count after scaling: {token}")
        return token
    else:
        raise ValueError("Image path cannot be empty. Provide a valid image file path.")
    
if __name__ == "__main__":
    token = token_calculate(image_path="xxx/test.jpg")

Video

Video file tokens are divided into video_tokens (visual) and audio_tokens (audio).

video_tokens

The calculation is complex. See the following code:

# Before use, install: pip install opencv-python
import math
import os
import logging
import cv2

# Fixed parameters
FRAME_FACTOR = 2

# For qwen3-omni-flash, IMAGE_FACTOR is 32
IMAGE_FACTOR = 32

# For qwen-omni-turbo, IMAGE_FACTOR is 28
# IMAGE_FACTOR = 28

# Video frame aspect ratio
MAX_RATIO = 200

# Minimum video frame pixels. For qwen3-omni-flash: 128 * 32 * 32
VIDEO_MIN_PIXELS = 128 * 32 * 32
# For qwen-omni-turbo
# VIDEO_MIN_PIXELS = 128 * 28 * 28

# Maximum video frame pixels. For qwen3-omni-flash: 768 * 32 * 32
VIDEO_MAX_PIXELS = 768 * 32 * 32
# For qwen-omni-turbo:
# VIDEO_MAX_PIXELS = 768 * 28 * 28

FPS = 2
# Minimum extracted frames
FPS_MIN_FRAMES = 4

# Maximum extracted frames
# Maximum extracted frames for qwen3-omni-flash: 128
# Maximum extracted frames for qwen-omni-turbo: 80
FPS_MAX_FRAMES = 128

# Maximum pixel value for video input. For qwen3-omni-flash: 16384 * 32 * 32
VIDEO_TOTAL_PIXELS = 16384 * 32 * 32
# For qwen-omni-turbo:
# VIDEO_TOTAL_PIXELS = 16384 * 28 * 28

def round_by_factor(number, factor):
    return round(number / factor) * factor

def ceil_by_factor(number, factor):
    return math.ceil(number / factor) * factor

def floor_by_factor(number, factor):
    return math.floor(number / factor) * factor

def get_video(video_path):
    cap = cv2.VideoCapture(video_path)
    frame_width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
    frame_height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
    total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
    video_fps = cap.get(cv2.CAP_PROP_FPS)
    cap.release()
    return frame_height, frame_width, total_frames, video_fps

def smart_nframes(total_frames, video_fps):
    min_frames = ceil_by_factor(FPS_MIN_FRAMES, FRAME_FACTOR)
    max_frames = floor_by_factor(min(FPS_MAX_FRAMES, total_frames), FRAME_FACTOR)
    duration = total_frames / video_fps if video_fps != 0 else 0
    if duration - int(duration) > (1 / FPS):
        total_frames = math.ceil(duration * video_fps)
    else:
        total_frames = math.ceil(int(duration) * video_fps)
    nframes = total_frames / video_fps * FPS
    nframes = int(min(min(max(nframes, min_frames), max_frames), total_frames))
    if not (FRAME_FACTOR <= nframes <= total_frames):
        raise ValueError(f"nframes should be in interval [{FRAME_FACTOR}, {total_frames}], but got {nframes}.")
    return nframes

def smart_resize(height, width, nframes, factor=IMAGE_FACTOR):
    min_pixels = VIDEO_MIN_PIXELS
    total_pixels = VIDEO_TOTAL_PIXELS
    max_pixels = max(min(VIDEO_MAX_PIXELS, total_pixels / nframes * FRAME_FACTOR), int(min_pixels * 1.05))
    if max(height, width) / min(height, width) > MAX_RATIO:
        raise ValueError(f"absolute aspect ratio must be smaller than {MAX_RATIO}, got {max(height, width) / min(height, width)}")
    h_bar = max(factor, round_by_factor(height, factor))
    w_bar = max(factor, round_by_factor(width, factor))
    if h_bar * w_bar > max_pixels:
        beta = math.sqrt((height * width) / max_pixels)
        h_bar = floor_by_factor(height / beta, factor)
        w_bar = floor_by_factor(width / beta, factor)
    elif h_bar * w_bar < min_pixels:
        beta = math.sqrt(min_pixels / (height * width))
        h_bar = ceil_by_factor(height * beta, factor)
        w_bar = ceil_by_factor(width * beta, factor)
    return h_bar, w_bar

def video_token_calculate(video_path):
    height, width, total_frames, video_fps = get_video(video_path)
    nframes = smart_nframes(total_frames, video_fps)
    resized_height, resized_width = smart_resize(height, width, nframes)
    video_token = int(math.ceil(nframes / FPS) * resized_height / 32 * resized_width / 32)
    video_token += 2  # Visual markers
    return video_token

if __name__ == "__main__":
    video_path = "spring_mountain.mp4"  # Your video path
    video_token = video_token_calculate(video_path)
    print("video_tokens:", video_token)

audio_tokens
- qwen3-omni-flash: Total tokens = Audio duration (seconds) × 12.5
- qwen-omni-turbo: Total tokens = Audio duration (seconds) × 25
Audio shorter than 1 second is billed as 1 second.

Free quota

For more information about how to claim, query, and use your free quota, see Free quota for new users.

Rate limits

For more information about model rate limit rules and frequently asked questions, see Rate limits.

Error codes

If the model call fails and returns an error message, see Error messages for resolution.

Voice list

When you use this feature, set the voice request parameter to the corresponding value in the "voice parameter" column of the table below:

qwen3-omni-flash-2025-12-01 model

Voice name	`voice` parameter	Voice effect	Description	Languages supported
Cherry	Cherry		A sunny, positive, friendly, and natural young woman	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Serena	Serena		A gentle young woman	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Ethan	Ethan		Standard Mandarin with a slight northern accent. Sunny, warm, energetic, and vibrant	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Chelsie	Chelsie		A two-dimensional virtual girlfriend	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Momo	Momo		Playful and mischievous, cheering you up	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Vivian	Vivian		Confident, cute, and slightly feisty	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Moon	Moon		Effortlessly cool Moon White	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Maia	Maia		A blend of intellect and gentleness	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Kai	Kai		A soothing audio spa for your ears	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Nofish	Nofish		A designer who cannot pronounce retroflex sounds	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Bella	Bella		A little girl who drinks but never throws punches when drunk	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Jennifer	Jennifer		A premium, cinematic-quality American English female voice	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Ryan	Ryan		Full of rhythm, bursting with dramatic flair, balancing authenticity and tension	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Katerina	Katerina		A mature-woman voice with rich, memorable rhythm	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Aiden	Aiden		An American English young man skilled in cooking	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Eldric Sage	Eldric Sage		A calm and wise elder—weathered like a pine tree, yet clear-minded as a mirror	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Mia	Mia		Gentle as spring water, obedient as fresh snow	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Mochi	Mochi		A clever, quick-witted young adult—childlike innocence remains, yet wisdom shines through	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Bellona	Bellona		A powerful, clear voice that brings characters to life—so stirring it makes your blood boil. With heroic grandeur and perfect diction, this voice captures the full spectrum of human expression.	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Vincent	Vincent		A uniquely raspy, smoky voice—just one line evokes armies and heroic tales	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Bunny	Bunny		A little girl overflowing with "cuteness"	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Neil	Neil		A flat baseline intonation with precise, clear pronunciation—the most professional news anchor	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Elias	Elias		Maintains academic rigor while using storytelling techniques to turn complex knowledge into digestible learning modules	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Arthur	Arthur		A simple, earthy voice steeped in time and tobacco smoke—slowly unfolding village stories and curiosities	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Nini	Nini		A soft, clingy voice like sweet rice cakes—those drawn-out calls of “Big Brother” are so sweet they melt your bones	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Ebona	Ebona		Her whisper is like a rusty key slowly turning in the darkest corner of your mind—where childhood shadows and unknown fears hide	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Seren	Seren		A gentle, soothing voice to help you fall asleep faster. Good night, sweet dreams	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Pip	Pip		A playful, mischievous boy full of childlike wonder—is this your memory of Shin-chan?	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Stella	Stella		Normally a cloyingly sweet, dazed teenage-girl voice—but when shouting “I represent the moon to defeat you!”, she instantly radiates unwavering love and justice	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Bodega	Bodega		A passionate Spanish man	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Sonrisa	Sonrisa		A cheerful, outgoing Latin American woman	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Alek	Alek		Cold like the Russian spirit, yet warm like wool coat lining	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Dolce	Dolce		A laid-back Italian man	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Sohee	Sohee		A warm, cheerful, emotionally expressive Korean unnie	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Ono Anna	Ono Anna		A clever, spirited childhood friend	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Lenn	Lenn		Rational at heart, rebellious in detail—a German youth who wears suits and listens to post-punk	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Emilien	Emilien		A romantic French big brother	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Andre	Andre		A magnetic, natural, and steady male voice	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Radio Gol	Radio Gol		Football poet Radio Gol! Today I’ll commentate on football using my name.	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Shanghai - Jada	Jada		A fast-paced, energetic Shanghai auntie	Shanghainese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Beijing - Dylan	Dylan		A young man raised in Beijing’s hutongs	Beijing dialect, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Nanjing - Li	Li		A patient yoga teacher	Nanjing dialect, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Shaanxi - Marcus	Marcus		Broad face, few words, sincere heart, deep voice—the authentic Shaanxi flavor	Shaanxi dialect, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Southern Min - Roy	Roy		A humorous, straightforward, lively Taiwanese guy	Southern Min, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Tianjin - Peter	Peter		Tianjin-style crosstalk, professional foil	Tianjin dialect, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Sichuan - Sunny	Sunny		A Sichuan girl sweet enough to melt your heart	Sichuan dialect, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Sichuan - Eric	Eric		A Sichuanese man from Chengdu who stands out in everyday life	Sichuan dialect, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Cantonese - Rocky	Rocky		A humorous, witty A Qiang providing live chat	Cantonese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Cantonese - Kiki	Kiki		A sweet Hong Kong girl best friend	Cantonese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean

qwen3-omni-flash and qwen3-omni-flash-2025-09-15 models

Voice name	`voice` parameter	Voice effect	Description	Languages supported
Cherry	Cherry		A sunny, positive, friendly, and natural young woman	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Ethan	Ethan		Standard Mandarin with a slight northern accent. Sunny, warm, energetic, and vibrant	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Nofish	Nofish		A designer who cannot pronounce retroflex sounds	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Jennifer	Jennifer		A premium, cinematic-quality American English female voice	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Ryan	Ryan		Full of rhythm, bursting with dramatic flair, balancing authenticity and tension	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Katerina	Katerina		A mature-woman voice with rich, memorable rhythm	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Elias	Elias		Maintains academic rigor while using storytelling techniques to turn complex knowledge into digestible learning modules	Chinese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Shanghai - Jada	Jada		A fast-paced, energetic Shanghai auntie	Shanghainese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Beijing - Dylan	Dylan		A young man raised in Beijing’s hutongs	Beijing dialect, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Sichuan - Sunny	Sunny		A Sichuan girl sweet enough to melt your heart	Sichuan dialect, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Nanjing - Li	Li		A patient yoga teacher	Nanjing dialect, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Shaanxi - Marcus	Marcus		Broad face, few words, sincere heart, deep voice—the authentic Shaanxi flavor	Shaanxi dialect, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Southern Min - Roy	Roy		A humorous, straightforward, lively Taiwanese guy	Southern Min, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Tianjin - Peter	Peter		Tianjin-style crosstalk, professional foil	Tianjin dialect, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Cantonese - Rocky	Rocky		A humorous, witty A Qiang providing live chat	Cantonese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Cantonese - Kiki	Kiki		A sweet Hong Kong girl best friend	Cantonese, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean
Sichuan - Eric	Eric		A Sichuanese man from Chengdu who stands out in everyday life	Sichuan dialect, English, French, German, Russian, Italian, Spanish, Portuguese, Japanese, Korean

Qwen-Omni-Turbo model

Voice name	`voice` parameter	Voice Effect	Description	Languages supported
Cherry	Cherry		A sunny, positive, friendly, and natural young woman	Chinese, English
Serena	Serena		A gentle young woman	Chinese, English
Ethan	Ethan		Standard Mandarin with a slight northern accent. Sunny, warm, energetic, and vibrant	Chinese, English
Chelsie	Chelsie		A two-dimensional virtual girlfriend	Chinese, English

Qwen-Omni open-source models

Voice name	`voice` parameter	Voice Effects	Description	Languages supported
Ethan	Ethan		Standard Mandarin with a slight northern accent. Sunny, warm, energetic, and vibrant	Chinese, English
Chelsie	Chelsie		A two-dimensional virtual girlfriend	Chinese, English