Non-real-time speech recognition (Qwen-Audio-ASR) HTTP API

For input requirements such as supported audio formats, file size limits, and duration limits, see Audio specifications.

Service endpoints

The following uses the Singapore (ap-southeast-1) endpoint. For another region, replace ap-southeast-1 with the region ID and use an API key from that region. See Regions and endpoints. POST https://ap-southeast-1.maas.qwencloudapi.com/api/v1/services/aigc/multimodal-generation/generation

Request headers

Parameter	Type	Required	Description
Authorization	string	Yes	The authentication token, in the format `Bearer $DASHSCOPE_API_KEY`. Replace "`$DASHSCOPE_API_KEY`" with your actual API key.
Content-Type	string	Yes	The media type of the request body. Fixed to `application/json`.
X-DashScope-SSE	string	No	Controls whether to request streaming results. Set to `enable` to request SSE streaming. Set to `disable` or omit this parameter to return only the final result. Applies only to `qwen-audio-3.0-asr-flash`: SSE streaming is available only for audio that is at least 1 minute long.

Request body

# Singapore (ap-southeast-1). For another region, replace ap-southeast-1 in the URL.
# API keys are region-specific. See https://docs.qwencloud.com/developer-guides/getting-started/regions
curl --location --request POST 'https://ap-southeast-1.maas.qwencloudapi.com/api/v1/services/aigc/multimodal-generation/generation' \
     --header "Authorization: Bearer $DASHSCOPE_API_KEY" \
     --header "Content-Type: application/json" \
     --header "X-DashScope-SSE: disable" \
     --data '{
    "model": "qwen-audio-3.1-asr-flash",
    "input": {
        "messages": [
            {
                "role": "user",
                "content": [
                    {
                        "type": "input_audio",
                        "input_audio": {
                            "data": "{YOUR_AUDIO_URL}"
                        }
                    }
                ]
            }
        ]
    },
    "parameters": {
        "format": "wav",
        "sample_rate": "16000"
    }
}'

# Singapore (ap-southeast-1). For another region, replace ap-southeast-1 in the URL.
# API keys are region-specific. See https://docs.qwencloud.com/developer-guides/getting-started/regions
curl --location --request POST 'https://ap-southeast-1.maas.qwencloudapi.com/api/v1/services/aigc/multimodal-generation/generation' \
     --header "Authorization: Bearer $DASHSCOPE_API_KEY" \
     --header "Content-Type: application/json" \
     --header "X-DashScope-SSE: enable" \
     --data '{
    "model": "qwen-audio-3.1-asr-flash",
    "input": {
        "messages": [
            {
                "role": "user",
                "content": [
                    {
                        "type": "input_audio",
                        "input_audio": {
                            "data": "{YOUR_AUDIO_URL}"
                        }
                    }
                ]
            }
        ]
    },
    "parameters": {
        "format": "wav",
        "sample_rate": "16000"
    }
}'

This original example contains an unescaped apostrophe inside a shell-quoted JSON string. Use the corrected heredoc example in the Shell quoting corrections section below.

# Singapore (ap-southeast-1). For another region, replace ap-southeast-1 in the URL.
# API keys are region-specific. See https://docs.qwencloud.com/developer-guides/getting-started/regions
curl --location --request POST 'https://ap-southeast-1.maas.qwencloudapi.com/api/v1/services/aigc/multimodal-generation/generation' \
     --header "Authorization: Bearer $DASHSCOPE_API_KEY" \
     --header "Content-Type: application/json" \
     --header "X-DashScope-SSE: disable" \
     --data '{
    "model": "qwen-audio-3.1-asr-flash",
    "input": {
        "messages": [
            {
                "role": "user",
                "content": [
                    {
                        "type": "input_text",
                        "text": "Hello"
                    }
                ]
            },
            {
                "role": "assistant",
                "content": [
                    {
                        "type": "text",
                        "text": "Hello, I'm Qwen. How can I help you?"
                    }
                ]
            },
            {
                "role": "user",
                "content": [
                    {
                        "type": "input_audio",
                        "input_audio": {
                            "data": "{YOUR_AUDIO_URL}"
                        }
                    }
                ]
            }
        ]
    },
    "parameters": {
        "format": "wav",
        "sample_rate": "16000"
    }
}'

This original example contains an unescaped apostrophe inside a shell-quoted JSON string. Use the corrected heredoc example in the Shell quoting corrections section below.

# Singapore (ap-southeast-1). For another region, replace ap-southeast-1 in the URL.
# API keys are region-specific. See https://docs.qwencloud.com/developer-guides/getting-started/regions
curl --location --request POST 'https://ap-southeast-1.maas.qwencloudapi.com/api/v1/services/aigc/multimodal-generation/generation' \
     --header "Authorization: Bearer $DASHSCOPE_API_KEY" \
     --header "Content-Type: application/json" \
     --header "X-DashScope-SSE: enable" \
     --data '{
    "model": "qwen-audio-3.1-asr-flash",
    "input": {
        "messages": [
            {
                "role": "user",
                "content": [
                    {
                        "type": "input_text",
                        "text": "Hello"
                    }
                ]
            },
            {
                "role": "assistant",
                "content": [
                    {
                        "type": "text",
                        "text": "Hello, I'm Qwen. How can I help you?"
                    }
                ]
            },
            {
                "role": "user",
                "content": [
                    {
                        "type": "input_audio",
                        "input_audio": {
                            "data": "{YOUR_AUDIO_URL}"
                        }
                    }
                ]
            }
        ]
    },
    "parameters": {
        "format": "wav",
        "sample_rate": "16000"
    }
}'

You can pass Base64-encoded data (Data URL) in the format data:<mediatype>;base64,<data>.

<mediatype>: The MIME type.

The value depends on the audio format. For example:

WAV: audio/wav
MP3: audio/mpeg
<data>: The Base64-encoded string of the audio.

Base64 encoding increases the data size. For audio file size and duration limits, see Audio specifications.

Example: data:audio/wav;base64,SUQzBAAAAAAAI1RTU0UAAAAPAAADTGF2ZjU4LjI5LjEwMAAAAAAAAAAAAAAA//PAxABQ/BXRbMPe4IQAhl9

import base64, pathlib

# Replace with the path to your own audio file and make sure it meets the audio requirements
file_path = pathlib.Path("{YOUR_AUDIO_FILE}")
base64_str = base64.b64encode(file_path.read_bytes()).decode()
data_uri = f"data:audio/mpeg;base64,{base64_str}"

import base64, pathlib
import os
import requests

# Replace with the path to your own audio file and make sure it meets the audio requirements
file_path = pathlib.Path("{YOUR_AUDIO_FILE}")
base64_str = base64.b64encode(file_path.read_bytes()).decode()
data_uri = f"data:audio/wav;base64,{base64_str}"

# Singapore (ap-southeast-1). For another region, replace ap-southeast-1 in the URL.
# API keys are region-specific. See https://docs.qwencloud.com/developer-guides/getting-started/regions
url = "https://ap-southeast-1.maas.qwencloudapi.com/api/v1/services/aigc/multimodal-generation/generation"

headers = {
    "Authorization": f"Bearer {os.environ['DASHSCOPE_API_KEY']}",
    "Content-Type": "application/json",
    "X-DashScope-SSE": "disable",
}

payload = {
    "model": "qwen-audio-3.1-asr-flash",
    "input": {
        "messages": [
            {
                "role": "user",
                "content": [
                    {
                        "type": "input_audio",
                        "input_audio": {
                            "data": data_uri,
                        },
                    }
                ],
            }
        ]
    },
    "parameters": {
        "format": "wav",
        "sample_rate": "16000",
    },
}

response = requests.post(url, headers=headers, json=payload)
print(response.status_code)
print(response.json())

# Singapore (ap-southeast-1). For another region, replace ap-southeast-1 in the URL.
# API keys are region-specific. See https://docs.qwencloud.com/developer-guides/getting-started/regions
curl --location --request POST 'https://ap-southeast-1.maas.qwencloudapi.com/api/v1/services/aigc/multimodal-generation/generation' \
     --header "Authorization: Bearer $DASHSCOPE_API_KEY" \
     --header "Content-Type: application/json" \
     --header "X-DashScope-SSE: disable" \
     --data '{
    "model": "qwen-audio-3.1-asr-flash",
    "input": {
        "messages": [
            {
                "role": "user",
                "content": [
                    {
                        "type": "input_audio",
                        "input_audio": {
                            "data": "{YOUR_AUDIO_URL}"
                        }
                    }
                ]
            }
        ]
    },
    "parameters": {
        "format": "wav",
        "sample_rate": "16000",
        "vocabulary": {"John Smith": 5, "Jane Doe": 5}
    }
}'

Parameter	Type	Description
model	string	(Required) The model name.
input	object	(Required) The input information.
input.messages	array(object)	(Required) The message list. It contains the audio to recognize and, optionally, conversation context that improves recognition accuracy. Tip: The context feature improves the recognition accuracy of domain-specific terms. For usage, see Context enhancement. Constraints: A request can include at most 5 context messages of each type (`input_text` and `text`). If you exceed this limit, only the 5 most recent messages are kept. The total context text per turn (the combined length of the `text` fields in the `user` and `assistant` messages) cannot exceed 400 characters, where each character counts as 1. Any excess is truncated from the end. Tip: When you include context, the message order in `messages` matters: context messages must be arranged by conversation turn. Within each turn, the `user` message (of type `input_text`) must come before the corresponding `assistant` message (of type `text`). The `user` message that contains `input_audio` must be the last item in the `messages` array.
input.messages.role	string	(Required) The message role. Valid values: `user` (Required): A user message. When type is `input_audio`, it holds the audio to recognize. When type is `input_text`, it holds recognition results from previous turns or a domain-specific word list (optional, used as context). `assistant` (Optional, context): Replies from the large language model in previous turns.
input.messages.content	array(object)	(Required) The message content list.
input.messages.content.type	string	(Required) The content type. Each request needs at least one message of type `input_audio`. Valid values: `input_audio` (Required): The audio input to recognize (role is user). You must also pass the `input_audio` object. `input_text` (Optional, context): Recognition results of the user's speech from previous turns, or a domain-specific word list (role is user). You must also pass the `text` field. `text` (Optional, context): Replies from the large language model in previous turns (role is assistant). You must also pass the `text` field.
input.messages.content.input_audio	object	(Conditionally required) Required when `type` is `input_audio`.
input.messages.content.input_audio.data	string	(Required) The audio data to recognize. For input requirements such as supported audio formats, file size limits, and duration limits, see Audio specifications. Two methods are supported: Audio file URL: Pass a publicly accessible URL to the audio file. Base64 Data URI: Pass the Base64-encoded audio data as a Data URI. The value is the `data:{MIME_TYPE};base64,` prefix concatenated with the Base64-encoded audio data. Supported MIME types include `audio/wav` and `audio/mp3`. Example (URL): `https://example.com/audio/sample.wav` Example (Base64): `data:audio/wav;base64,{BASE64_ENCODED_DATA}`
input.messages.content.text	string	(Conditionally required) When `type` is `input_text`, enter the recognition results of the user's speech from previous turns or a domain-specific word list. When `type` is `text`, enter the large language model's replies from previous turns. Text length is measured in characters, where each character counts as 1. The combined length of the `text` fields across all messages in a single context turn cannot exceed 400 characters. Any excess is truncated from the end.
parameters	object	(Required) The model parameters. Note: Text polishing is disabled by default and is not yet available for `qwen-audio-3.0-asr-flash`. `qwen-audio-3.1-asr-flash` supports native text polishing. Text Polishing: While transcribing speech, the model automatically removes meaningless filler words and stuttered repetitions, handles self-corrections made during speech, smooths out colloquial expressions, and standardizes punctuation and text formatting. This results in output that is more concise, fluent, and readable, while preserving the user's original intent and key information as much as possible.
parameters.speaker_diarization_enabled	boolean	(Optional) Supported only by `qwen-audio-3.1-asr-flash`. Specifies whether to enable speaker diarization. Default: `false`. When set to `true`, the response uses `speaker_id` to distinguish speakers and includes `output.sentences`. When diarization is enabled, `keep_dialect` and `language_hints` are ignored.
parameters.keep_dialect	boolean	(Optional) Supported only by `qwen-audio-3.1-asr-flash`. Specifies whether to preserve dialect expressions. Default: `false`. Set to `false` to transcribe dialect speech as standard Mandarin text, or `true` to preserve dialect expressions.
parameters.format	string	(Required) The audio format. Set this to match your actual audio format. Supported values include `wav`, `mp3`, and `opus`. For details, see Audio specifications.
parameters.sample_rate	string	(Optional) The audio sample rate, in Hz. For example, `16000` means a 16 kHz sample rate. For details, see Audio specifications.
parameters.vocabulary_id	string	(Optional) The ID of a precompiled hot word list. Generate this ID in advance by calling the create hot word list API. Pass the ID during recognition to use the hot words in the list. Suitable for scenarios where the vocabulary is known and relatively stable, and where you need to reuse the same word list across requests. For usage details, see Precompiled hotwords.
parameters.vocabulary	object	(Optional) Instant hot words. Passed as key-value pairs, where the key is the hot word text (`string`) and the value is the hot word weight (`integer`). No hot word list needs to be created in advance. The weight ranges from [1, 5] or is set to 50: a value in [1, 5] makes the model more likely to output the word as the value increases; a value of 50 designates a super hot word, which greatly improves recall, but the number of super hot words cannot exceed 50. Suitable for temporary, session-level hot word optimization. When instant and precompiled hotwords are configured together, the system merges both sets. If the merged set contains more than 2000 hotwords, the system randomly selects 2000 to use. For usage details, see Instant hotwords. Tip: Only `qwen-audio-3.1-asr-flash` and `qwen-audio-3.0-asr-flash` support inline hotwords.
parameters.language_hints	array[string]	(Optional) The language codes to recognize. If you cannot determine the language in advance, leave this unset and the model detects the language automatically. You can set up to 4 values. If you set more, only the first 4 take effect. qwen-audio-3.1-asr-flash, qwen-audio-3.0-asr-flash: zh: Chinese en: English ja: Japanese ko: Korean vi: Vietnamese th: Thai id: Indonesian ms: Malay tl: Filipino hi: Hindi ar: Arabic fr: French de: German es: Spanish pt: Portuguese ru: Russian it: Italian nl: Dutch sv: Swedish da: Danish fi: Finnish no: Norwegian el: Greek pl: Polish cs: Czech hu: Hungarian ro: Romanian bg: Bulgarian hr: Croatian sk: Slovak Specifying language codes can improve recognition accuracy. qwen-audio-3.1-asr-flash also supports the following language codes: tr: Turkish fa: Persian mr: Marathi ml: Malayalam ne: Nepali ky: Kyrgyz az: Azerbaijani ka: Georgian hy: Armenian sl: Slovenian sq: Albanian et: Estonian lv: Latvian lt: Lithuanian ca: Catalan eu: Basque gl: Galician or: Odia dv: Divehi as: Assamese ku: Kurdish tig: Tigre fy: Western Frisian ba: Bashkir be: Belarusian tt: Tatar ia: Interlingua mhr: Eastern Mari tok: Toki Pona

Response body

Non-streaming
Streaming

{
    "output": {
        "sentence": {
            "begin_time": 760,
            "channel_id": 0,
            "end_time": 3800,
            "sentence_end": true,
            "sentence_id": 1,
            "text": "Hello World, this is Alibaba Speech Lab.",
            "words": [
                {"begin_time": 760, "end_time": 1040, "fixed": true, "punctuation": "", "text": "Hello"},
                {"begin_time": 1040, "end_time": 1240, "fixed": true, "punctuation": ",", "text": " World"},
                {"begin_time": 1360, "end_time": 1880, "fixed": true, "punctuation": "", "text": "this is"},
                {"begin_time": 1880, "end_time": 2520, "fixed": true, "punctuation": "", "text": "Alibaba"},
                {"begin_time": 2520, "end_time": 2840, "fixed": true, "punctuation": "", "text": "Speech"},
                {"begin_time": 2840, "end_time": 3800, "fixed": true, "punctuation": ".", "text": "Lab"}
            ]
        },
        "text": "Hello World, this is Alibaba Speech Lab."
    },
    "usage": {
        "duration": 4
    },
    "request_id": "40e0734d-096f-9ae3-86c1-a8c013287561"
}

Only when the audio is at least 1 minute long and X-DashScope-SSE: enable is set does the server return recognition results using the Server-Sent Events protocol. The SSE event format is as follows:

id:{sequence_number}
      event:result
      :HTTP_STATUS/200
      data:{JSON_data}

Example response:

id:1
event:result
:HTTP_STATUS/200
data:{"output":{"sentence":{"sentence_id":1,"sentence_end":true,"end_time":3800,"words":[{"end_time":1040,"punctuation":"","begin_time":760,"fixed":true,"text":"Hello"},{"end_time":1240,"punctuation":",","begin_time":1040,"fixed":true,"text":" World"},{"end_time":1880,"punctuation":"","begin_time":1360,"fixed":true,"text":"this is"},{"end_time":2520,"punctuation":"","begin_time":1880,"fixed":true,"text":"Alibaba"},{"end_time":2840,"punctuation":"","begin_time":2520,"fixed":true,"text":"Speech"},{"end_time":3800,"punctuation":".","begin_time":2840,"fixed":true,"text":"Lab"}],"begin_time":760,"text":"Hello World, this is Alibaba Speech Lab.","channel_id":0},"text":"Hello World, this is Alibaba Speech Lab."},"usage":{"duration":4},"request_id":"fc1582e4-935c-9fc2-a482-a98bf43daa69"}

Parameter	Type	Description
request_id	string	The unique identifier of this request.
output	object	The output result.
output.text	string	The full recognized text accumulated so far.
output.sentences	array[object]	Returned only when speaker diarization is enabled for `qwen-audio-3.1-asr-flash`. Contains recognition results for the current and previous sentences. Each element has the same structure as `sentence`.
output.sentence	object	The details of the current sentence.
output.sentence.speaker_id	integer \| null	The speaker ID, used to distinguish speakers. For `qwen-audio-3.1-asr-flash`, this contains an ID when speaker diarization is enabled, and is `null` otherwise.
output.sentence.sentence_id	integer	The sentence number, starting from 1.
output.sentence.sentence_end	boolean	Whether this is the final result for the sentence. `true` indicates that recognition of the sentence is complete.
output.sentence.begin_time	integer	The start time of the sentence, in milliseconds.
output.sentence.end_time	integer	The end time of the sentence, in milliseconds. Returned only when `sentence_end` is `true`.
output.sentence.text	string	The recognized text of the current sentence.
output.sentence.channel_id	integer	The channel number, starting from 0.
output.sentence.words	array	The list of word-level timestamps.
output.sentence.words.speaker_id	integer \| null	The ID of the speaker for this word. For `qwen-audio-3.1-asr-flash`, this contains an ID when speaker diarization is enabled, and is `null` otherwise.
output.sentence.words.text	string	The word text.
output.sentence.words.begin_time	integer	The start time of the word, in milliseconds.
output.sentence.words.end_time	integer	The end time of the word, in milliseconds.
output.sentence.words.punctuation	string	The punctuation mark after the word. An empty string when there is no punctuation.
output.sentence.words.fixed	boolean	Whether the word is stabilized. `false` indicates that the word's timestamp may be adjusted in subsequent events.
usage	object	The usage information. Returned only when `sentence_end` is `true`.
usage.duration	integer	The processed audio duration, in seconds.
usage.input_tokens	integer	The number of input tokens consumed by this request. Applies only to `qwen-audio-3.1-asr-flash`.
usage.output_tokens	integer	The number of output tokens consumed by this request. Applies only to `qwen-audio-3.1-asr-flash`.
usage.total_tokens	integer	The total number of tokens consumed by this request, equal to the sum of `input_tokens` and `output_tokens`. Applies only to `qwen-audio-3.1-asr-flash`.

SSE streaming result processing logic

In streaming mode, the client needs to handle the following:

For each SSE event received, parse the JSON in the data field.
For qwen-audio-3.0-asr-flash, use output.sentence.sentence_end to determine whether the current sentence has ended. When this value is true, recognition of the sentence is complete, the word-level timestamps are stabilized, and the result can be used as final. When this value is false, recognition is still in progress, and the text and timestamps may be updated in subsequent events.
The usage information is returned only in the sentence-end event and includes the processed audio duration. For qwen-audio-3.1-asr-flash, it also includes the input, output, and total token counts for the request. This model is billed by token usage.

For qwen-audio-3.1-asr-flash without speaker diarization, a word timestamp becomes final only when output.sentence.words[i].fixed is true. Until then, timestamps may change. When speaker diarization is enabled, read the current and previous sentence results from output.sentences.

Shell quoting corrections

These examples preserve all request values and use a quoted heredoc so apostrophes in the JSON text are passed literally.

Corrected non-streaming request

# Singapore (ap-southeast-1). For another region, replace ap-southeast-1 in the URL.
# API keys are region-specific. See https://docs.qwencloud.com/developer-guides/getting-started/regions
curl --location --request POST 'https://ap-southeast-1.maas.qwencloudapi.com/api/v1/services/aigc/multimodal-generation/generation' \
     --header "Authorization: Bearer $DASHSCOPE_API_KEY" \
     --header "Content-Type: application/json" \
     --header "X-DashScope-SSE: disable" \
     --data @- <<'JSON'
{
    "model": "qwen-audio-3.1-asr-flash",
    "input": {
        "messages": [
            {
                "role": "user",
                "content": [
                    {
                        "type": "input_text",
                        "text": "Hello"
                    }
                ]
            },
            {
                "role": "assistant",
                "content": [
                    {
                        "type": "text",
                        "text": "Hello, I'm Qwen. How can I help you?"
                    }
                ]
            },
            {
                "role": "user",
                "content": [
                    {
                        "type": "input_audio",
                        "input_audio": {
                            "data": "{YOUR_AUDIO_URL}"
                        }
                    }
                ]
            }
        ]
    },
    "parameters": {
        "format": "wav",
        "sample_rate": "16000"
    }
}
JSON

Corrected streaming request

# Singapore (ap-southeast-1). For another region, replace ap-southeast-1 in the URL.
# API keys are region-specific. See https://docs.qwencloud.com/developer-guides/getting-started/regions
curl --location --request POST 'https://ap-southeast-1.maas.qwencloudapi.com/api/v1/services/aigc/multimodal-generation/generation' \
     --header "Authorization: Bearer $DASHSCOPE_API_KEY" \
     --header "Content-Type: application/json" \
     --header "X-DashScope-SSE: enable" \
     --data @- <<'JSON'
{
    "model": "qwen-audio-3.1-asr-flash",
    "input": {
        "messages": [
            {
                "role": "user",
                "content": [
                    {
                        "type": "input_text",
                        "text": "Hello"
                    }
                ]
            },
            {
                "role": "assistant",
                "content": [
                    {
                        "type": "text",
                        "text": "Hello, I'm Qwen. How can I help you?"
                    }
                ]
            },
            {
                "role": "user",
                "content": [
                    {
                        "type": "input_audio",
                        "input_audio": {
                            "data": "{YOUR_AUDIO_URL}"
                        }
                    }
                ]
            }
        ]
    },
    "parameters": {
        "format": "wav",
        "sample_rate": "16000"
    }
}
JSON

​Service endpoints

​Request headers

​Request body

​Response body

​SSE streaming result processing logic

​Shell quoting corrections

​Corrected non-streaming request

​Corrected streaming request

Service endpoints

Request headers

Request body

Response body

SSE streaming result processing logic

Shell quoting corrections

Corrected non-streaming request

Corrected streaming request