API reference for recording file recognition-Intelligent Speech Interaction(ISI)-阿里云帮助中心

Billing and service versions

Recording file recognition provides a trial edition and a commercial edition. For billing rules, see Billing.

For instructions on upgrading to the commercial edition and details about fees, see Service management and activation.

For new integrations, explicitly set version="4.0" in the task parameters. This version uses the same camel-case fields for polling and callback results. For the legacy callback format, see Legacy version compatibility.

Limits

File requirements

Single-track and dual-track WAV, MP3, MP4, M4A, WMA, AAC, OGG, AMR, and FLAC files are supported. Audio files must not exceed 512 MB, video files must not exceed 2 GB, and the total file duration must not exceed 12 hours.

The language and sample rate of the audio must match the ASR model configured for the project. For model selection and supported models, see Languages and models.

Store the recording file on a file server that the service can access. Use a domain name rather than an IP address as the URL host. URLs must not contain spaces. Avoid Chinese characters in URLs.

We recommend storing files in Alibaba Cloud OSS. For public-read files, use an accessible file URL. For private files, use a presigned URL that remains valid while the service downloads the file. For instructions, see Use presigned URLs to download or preview files.

For a self-hosted file server, ensure that the HTTP response header Content-Length matches the actual response body length.

Valid URL	Invalid URL
https://gw.alipayobjects.com/os/bmw-prod/0574ee2e-f494-45a5-820f-63aee583045a.wav	http://127.0.0.1/sample.wav D:\files\sample.wav

Access to OSS over an internal endpoint in the same region does not incur Internet traffic fees. For configuration, see Access recording files over an OSS internal endpoint.

Request rates and task retention

Recording file recognition is an offline service and does not limit the number of concurrent recognition tasks. API request rate limits are as follows:

Operation	User-level QPS limit
SubmitTask (POST)	200
GetTaskResult (GET)	500
GetTaskResult, same TaskId	1

Requests that exceed the rate limits return Throttling.User. Set a polling interval to control the request rate.

For more information, see Concurrency and request limits.

The trial period for new users is three months. Two hours of recording files can be recognized free of charge every 24 hours. After the quota is exhausted, another trial quota is available after 24 hours.

After submission, tasks for free users complete within 24 hours, and tasks for paid users complete within three hours. Recognition results are retained on the server for 72 hours.

Important

These processing times do not apply to large batches, such as more than 500 hours of recordings submitted within half an hour. Contact a presales specialist for large-scale recognition requirements.

Workflow

Configure the ASR model for the project based on the language and sample rate of the audio.
Prepare a downloadable URL for the recording file.
Call SubmitTask by using a POST request and pass the task settings in the Task parameter. A successful submission returns the task ID in the top-level TaskId field of the JSON response.
Obtain the result: for polling, call GetTaskResult with this TaskId; for callbacks, set enable_callback=true and callback_url. The service sends the result to the callback URL by using POST when the task completes.

Interaction flow

The following diagram shows the interaction between the client and server. The task ID is returned in the TaskId field of the JSON response.

Client-server interaction flow for recording file recognition

API configuration

The service provides RPC-style POP APIs authenticated by AccessKey signatures. SubmitTask uses POST, and GetTaskResult uses GET. The product name used in SDK configuration is nls-filetrans. Configure the region, endpoint, and API date version as follows:

Region	regionId / endpointName	domain	API date version (Version)
China (Shanghai)	cn-shanghai	filetrans.cn-shanghai.aliyuncs.com	2018-08-17
China (Beijing)	cn-beijing	filetrans.cn-beijing.aliyuncs.com	2018-08-17
China (Shenzhen)	cn-shenzhen	filetrans.cn-shenzhen.aliyuncs.com	2018-08-17

The API date version Version configures the POP request. The lowercase version in the task parameters selects the recording file recognition service version.

Request parameters

Submit a task: SubmitTask

Serialize the following JSON object as a string and pass it as the Task parameter in a POST request. In this example, valid_times limits recognition to the specified interval. Omit this field to recognize the entire file.

{
    "appkey": "your-appkey",
    "file_link": "https://gw.alipayobjects.com/os/bmw-prod/0574ee2e-f494-45a5-820f-63aee583045a.wav",
    "auto_split":false,
    "version": "4.0",
    "enable_words": false,
    "enable_sample_rate_adaptive": true,
    "valid_times": [
        {
            "begin_time": 200,
            "end_time":2000,
            "channel_id": 0
        }
    ]
}

Task fields

Parameter	Type	Required	Description
appkey	String	Yes	The Appkey of the project. The ASR model configured for the project must match the language and sample rate of the audio.
file_link	String	Yes	The HTTP or HTTPS download URL of the recording file. The service must be able to download the file from this URL.
version	String	No	The service version used by the task. For new integrations, explicitly set this parameter to 4.0. For callback format compatibility, see Legacy version compatibility.
enable_words	Boolean	No	Specifies whether to return word information. Default: false. Set version to 4.0 to enable this feature.
enable_sample_rate_adaptive	Boolean	No	Specifies whether to automatically downsample audio with a sample rate above 16 kHz to 16 kHz. Default: false. Set version to 4.0 to enable this feature.
enable_callback	Boolean	No	Specifies whether to enable callbacks. Default: false.
callback_url	String	No	The callback URL. Required when enable_callback is true. HTTP and HTTPS are supported. Use a domain name rather than an IP address as the host. The callback service must accept POST requests.
auto_split	Boolean	No	Specifies whether to enable speaker diarization. ChannelId in each sentence identifies its speaker. The first speaker usually has ChannelId 0. Both 8 kHz and 16 kHz audio are supported. For 8 kHz dual-channel audio, two speakers are used by default, and channel0 and channel1 identify the audio tracks. For 16 kHz audio, only the first channel is diarized by default.
supervise_type	Integer	No	How the number of speakers is determined. Use this parameter with auto_split and speaker_num. If omitted, the user specifies the number for 8 kHz audio, and the algorithm determines the number for 16 kHz audio. 1 uses the number specified by speaker_num; 2 lets the algorithm determine the number.
speaker_num	Integer	No	The expected number of speakers, used as a hint for the algorithm. Valid values: integers from 2 to 100. Default: 2 for 8 kHz audio and 100 for 16 kHz audio. Use this parameter with auto_split and supervise_type. The algorithm attempts to use this number, but the number of speakers in the result is not guaranteed to match it.
enable_inverse_text_normalization	Boolean	No	Specifies whether to enable inverse text normalization (ITN) to convert Chinese numerals to Arabic numerals. Default: false. Set version to 4.0 to enable this feature.
enable_disfluency	Boolean	No	Specifies whether to filter filler words. Default: false. Set version to 4.0 to enable this feature.
enable_punctuation_prediction	Boolean	No	Specifies whether to add punctuation to recognition results. Default: true.
valid_times	List< ValidTime >	No	A list of valid time intervals. Only the specified audio segments are recognized. For the fields of each element, see ValidTime fields below.
max_end_silence	Integer	No	The maximum end-of-sentence silence. Valid values: 200 to 6000 ms. Default: 800 ms. This parameter has no effect when enable_semantic_sentence_detection is enabled.
max_single_segment_time	Integer	No	The maximum duration of a sentence. Minimum: 5000 ms. Default: 60000 ms. This parameter has no effect when enable_semantic_sentence_detection is enabled.
customization_id	String	No	The ID of the custom language model created by using the POP API. Omitted by default.
class_vocabulary_id	String	No	The ID of the class vocabulary. Omitted by default.
vocabulary_id	String	No	The ID of the general vocabulary. Omitted by default.
enable_semantic_sentence_detection	Boolean	No	Specifies whether to enable semantic sentence segmentation. Valid values: true and false. Default: false.
enable_timestamp_alignment	Boolean	No	Specifies whether to enable timestamp alignment. Valid values: true and false. Default: false.
first_channel_only	Boolean	No	Specifies whether to recognize only the first channel. If omitted, both channels are processed for 8 kHz audio, and only the first channel is processed for 16 kHz audio. false processes both channels at either sample rate; true processes only the first channel. For 8 kHz dual-channel audio, billing is based on audio duration. For 16 kHz dual-channel audio, billing is based on the number of channels multiplied by audio duration. Set this parameter to true if the two channels contain duplicate content.
special_word_filter	String	No	Sensitive-word filtering settings. Custom words are supported. No filtering is performed by default. Specified words can be removed or replaced with *. For configuration, see the custom word filter example. The default word list is available at: Chinese sensitive-word list
punctuation_mark	String	No	Additional punctuation marks used for sentence segmentation. By default, periods, question marks, and exclamation marks are used. Specified marks are added to the defaults. For example, use "," for an English comma or "，," for both Chinese and English commas. Multiple punctuation marks are supported. Chinese and English punctuation are distinct. Do not insert spaces between marks. Non-punctuation characters have no effect.
sentence_max_length	Integer	No	The maximum number of characters displayed per sentence. Valid values: 4 to 50. Disabled by default. If enabled without a specified character count, long sentences are segmented. This parameter can be used to limit the number of characters per subtitle line.

ValidTime fields

Each element of valid_times contains the following fields:

Parameter	Type	Required	Description
begin_time	Int	Yes	The start offset of the valid time range. Unit: milliseconds.
end_time	Int	Yes	The end offset of the valid time range. Unit: milliseconds.
channel_id	Int	Yes	The track index to which the valid time range applies. The index starts from 0.

Custom word filter

Set special_word_filter in Task. system_reserved_filter=true enables filtering by the default word list. filter_with_empty.word_list specifies words to remove, and filter_with_signed.word_list specifies words to replace with *. These settings can be used individually or together.

{
    "appkey": "your-appkey",
    "file_link": "https://gw.alipayobjects.com/os/bmw-prod/0574ee2e-f494-45a5-820f-63aee583045a.wav",
    "version": "4.0",
    "special_word_filter": "{\"system_reserved_filter\":true,\"filter_with_empty\":{\"word_list\":[\"开始\",\"发生\"]},\"filter_with_signed\":{\"word_list\":[\"北京\"]}}"
}

Query a result: GetTaskResult

Pass the task ID returned by submission as the TaskId parameter in a GET request. Set the query interval according to the request rate limits.

Parameter	Type	Required	Description
TaskId	String	Yes	The recognition task ID.

Response parameters

The HTTP status indicates how the API request was processed. The task recognition status is reported by StatusCode and StatusText in the JSON response.

Submission response

A successful submission returns the following JSON:

{
        "TaskId": "4b56f0c4b7e611e88f34c33c2a60****",
        "RequestId": "E4B183CC-6CFE-411E-A547-D877F7BD****",
        "StatusText": "SUCCESS",
        "StatusCode": 21050000
}

Parameter	Type	Required	Description
TaskId	String	Yes	The recognition task ID.
RequestId	String	Yes	The request ID, used only for integration testing.
StatusCode	Int	Yes	The status code.
StatusText	String	Yes	The status description.

Query response

Succeeded

The following example shows the recognition result of the single-track Chinese recording nls-sample-16k.wav.

{
        "TaskId": "d429dd7dd75711e89305ab6170fe****",
        "RequestId": "9240D669-6485-4DCC-896A-F8B31F94****",
        "StatusText": "SUCCESS",
        "BizDuration": 2956,
        "SolveTime": 1540363288472,
        "StatusCode": 21050000,
        "Result": {
                "Sentences": [{
                        "EndTime": 2365,
                        "SilenceDuration": 0,
                        "BeginTime": 340,
                        "Text": "北京的天气。",
                        "ChannelId": 0,
                        "SpeechRate": 177,
                        "EmotionValue": 5.0
                }]
        }
}

Queued

Recognition has not started. Wait before querying again.

{
        "TaskId": "c7274235b7e611e88f34c33c2a60****",
        "RequestId": "981AD922-0655-46B0-8C6A-5C836822****",
        "StatusText": "QUEUEING",
        "StatusCode": 21050002
}

Running

Recognition is in progress. Wait before querying again.

{
        "TaskId": "c7274235b7e611e88f34c33c2a60****",
        "RequestId": "8E908ED2-867F-457E-82BF-4756194A****",
        "StatusText": "RUNNING",
        "BizDuration": 0,
        "StatusCode": 21050001
}

Failed

The following example shows a file download failure. Use StatusCode and StatusText to handle the error.

{
        "TaskId": "4cf25b7eb7e711e88f34c33c2a60****",
        "RequestId": "098BF27C-4CBA-45FF-BD11-3F532F26****",
        "StatusText": "FILE_DOWNLOAD_FAILED",
        "BizDuration": 0,
        "SolveTime": 1536906469146,
        "StatusCode": 41050002
}

Response fields

Sentences and Words are contained in the Result object.

Parameter	Type	Required	Description
TaskId	String	Yes	The recognition task ID.
StatusCode	Int	Yes	The status code.
StatusText	String	Yes	The status description.
RequestId	String	Yes	The request ID, used for debugging.
Result	Object	No	The recognition result object returned when the task succeeds. Not included in queued, running, or failed task responses.
Sentences	List< SentenceResult >	No	The sentence results in Result, returned when StatusText is SUCCESS.
Words	List< WordResult >	No	The word information in Result. Returned when enable_words is true and version is 4.0.
BizDuration	Long	No	The duration of the recognized audio, in milliseconds.
SolveTime	Long	No	The Unix timestamp when recognition completes, in milliseconds. Not returned before the task completes.

Sentences fields

Parameter	Type	Required	Description
ChannelId	Int	Yes	The track ID of the sentence.
BeginTime	Int	Yes	The start offset of the sentence. Unit: milliseconds.
EndTime	Int	Yes	The end offset of the sentence. Unit: milliseconds.
Text	String	Yes	The recognized text of the sentence.
EmotionValue	Float	Yes	The emotion energy value, calculated as the volume in decibels divided by 10. Valid range: 1 to 10.
SilenceDuration	Int	Yes	The silence duration between this sentence and the previous sentence. Unit: seconds.
SpeechRate	Int	Yes	The average speech rate of the sentence. For Chinese, the unit is characters per minute. For English, the unit is words per minute.

Callback result

When enable_callback=true, callback_url, and version="4.0" are set, the service sends a POST request to the callback URL when the task completes. The recognition result fields are the same as those returned by polling:

{
        "Result": {
                "Sentences": [{
                        "EndTime": 2365,
                        "SilenceDuration": 0,
                        "BeginTime": 340,
                        "Text": "北京的天气。",
                        "ChannelId": 0,
                        "SpeechRate": 177,
                        "EmotionValue": 5.0
                }]
        },
        "TaskId": "36d01b244ad811e9952db7bb7ed2****",
        "StatusCode": 21050000,
        "StatusText": "SUCCESS",
        "RequestTime": 1553062810452,
        "SolveTime": 1553062810831,
        "BizDuration": 2956
}

RequestTime is the Unix timestamp of task submission, and SolveTime is the Unix timestamp when recognition completes. Both are in milliseconds.

Word information

When enable_words=true and version="4.0" are set, the result includes Words. Polling and callbacks return the same word information. The following example shows the result of a Chinese recording:

{
        "StatusCode": 21050000,
        "Result": {
                "Sentences": [{
                        "SilenceDuration": 0,
                        "EmotionValue": 5.0,
                        "ChannelId": 0,
                        "Text": "北京的天气。",
                        "BeginTime": 340,
                        "EndTime": 2365,
                        "SpeechRate": 177
                }],
                "Words": [{
                        "ChannelId": 0,
                        "Word": "北京",
                        "BeginTime": 640,
                        "EndTime": 940
                }, {
                        "ChannelId": 0,
                        "Word": "的",
                        "BeginTime": 940,
                        "EndTime": 1120
                }, {
                        "ChannelId": 0,
                        "Word": "天气",
                        "BeginTime": 1120,
                        "EndTime": 2020
                }]
        },
        "SolveTime": 1553236968873,
        "StatusText": "SUCCESS",
        "RequestId": "027B126B-4AC8-4C98-9FEC-A031158F****",
        "TaskId": "b505e78c4c6d11e9a213e11db149****",
        "BizDuration": 2956
}

Words fields

Parameter	Type	Required	Description
BeginTime	Int	Yes	The start time of the word. Unit: milliseconds.
EndTime	Int	Yes	The end time of the word. Unit: milliseconds.
ChannelId	Int	Yes	The track ID of the word.
Word	String	Yes	The word text.

Task status and error handling

StatusCode	StatusText	Description
21050000	SUCCESS	SubmitTask: the task was submitted successfully. GetTaskResult: recognition succeeded.
21050001	RUNNING	Recognition is in progress. Wait before querying again.
21050002	QUEUEING	The task is queued. Wait before querying again.
21050003	SUCCESS_WITH_NO_VALID_FRAGMENT	The task completed, but no valid speech segment was detected. Check whether the file contains valid speech.

For error codes and solutions, see the recording file recognition errors in Error code reference.

Languages and models

Configure language and dialect models for the project. They cannot be selected through the task parameters of this API. Custom language models and hotwords are supported. For more information, see Language model customization and Hotwords.

On the All Projects page of the Intelligent Speech Interaction console, select the target project and open Project Settings. Select a model based on the language and sample rate. For instructions, see Manage projects.

Languages

Language	Model name	Sample rate	Punctuation	ITN	Disfluency removal	Semantic sentence detection	Audio-text alignment
English	General English, English for online education, and English for education content analysis	16k	Supported	Supported	Supported	Not supported	Supported
	Call center (general)	8k	Supported	Supported	Supported	Not supported	Not supported
	Southeast Asian languages	16k	Supported	Not supported	Not supported	Not supported	Not supported
Japanese	General - Japanese	16k	Supported	Supported	Not supported	Not supported	Supported
Spanish	General - Spanish	16k	Supported	Supported	Not supported	Not supported	Not supported
Spanish	Spanish for call centers	8k	Supported	Supported	Not supported	Not supported	Not supported
Arabic	General - Arabic	16k	Supported	Not supported	Not supported	Not supported	Not supported
Kazakh	General - Kazakh	16k	Supported	Not supported	Not supported	Not supported	Not supported
Korean	General - Korean	16k	Supported	Supported	Not supported	Not supported	Not supported
Thai	General - Thai	16k	Not supported	Not supported	Not supported	Not supported	Not supported
	Thai for call centers	8k	Not supported	Not supported	Not supported	Not supported	Not supported
	Southeast Asian languages	16k	Supported	Not supported	Not supported	Not supported	Not supported
Indonesian	General - Indonesian	16k	Supported	Supported	Not supported	Not supported	Not supported
	Call center (general)	8k	Supported	Supported	Not supported	Not supported	Not supported
	Southeast Asian languages	16k	Supported	Not supported	Not supported	Not supported	Not supported
Russian	General - Russian	16k	Supported	Supported	Not supported	Not supported	Not supported
Vietnamese	General - Vietnamese	16k	Supported	Supported	Not supported	Not supported	Not supported
	Vietnamese for call centers	8k	Supported	Supported	Not supported	Not supported	Not supported
	Southeast Asian languages	16k	Supported	Not supported	Not supported	Not supported	Not supported
French	General - French	16k	Supported	Supported	Not supported	Not supported	Not supported
German	General - German	16k	Supported	Supported	Not supported	Not supported	Not supported
Italian	General - Italian	16k	Supported	Not supported	Not supported	Not supported	Not supported
Hindi	General - Hindi	16k	Supported	Not supported	Not supported	Not supported	Not supported
Malay	General - Malay	16k	Supported	Not supported	Not supported	Not supported	Not supported
	Malay for call centers	8k	Supported	Not supported	Not supported	Not supported	Not supported
	Southeast Asian languages	16k	Supported	Not supported	Not supported	Not supported	Not supported
Filipino	General - Filipino	16k	Supported	Supported	Not supported	Not supported	Not supported
	Call center (general)	8k	Supported	Supported	Not supported	Not supported	Not supported
	Southeast Asian languages	16k	Supported	Not supported	Not supported	Not supported	Not supported
Tamil	General - Tamil	16k	Supported	Not supported	Not supported	Not supported	Not supported
Portuguese	General - Portuguese	16k	Supported	Supported	Not supported	Not supported	Not supported
Turkish	General - Turkish	16k	Supported	Not supported	Not supported	Not supported	Not supported
Polish	General - Polish	16k	Supported	Not supported	Not supported	Not supported	Not supported
Ukrainian	General - Ukrainian	16k	Supported	Not supported	Not supported	Not supported	Not supported
Romanian	General - Romanian	16k	Supported	Not supported	Not supported	Not supported	Not supported
Dutch	General - Dutch	16k	Supported	Not supported	Not supported	Not supported	Not supported
Greek	General - Greek	16k	Supported	Not supported	Not supported	Not supported	Not supported
Hungarian	General - Hungarian	16k	Supported	Not supported	Not supported	Not supported	Not supported
Javanese	General - Javanese	16k	Supported	Not supported	Not supported	Not supported	Not supported
Bengali	General - Bengali	16k	Supported	Not supported	Not supported	Not supported	Not supported
Burmese	General - Burmese	16k	Supported	Not supported	Not supported	Not supported	Not supported
Lao	General - Lao	16k	Supported	Not supported	Not supported	Not supported	Not supported
Swahili	General - Swahili	16k	Supported	Not supported	Not supported	Not supported	Not supported
Azerbaijani	General - Azerbaijani	16k	Supported	Not supported	Not supported	Not supported	Not supported
Persian	General - Persian	16k	Supported	Not supported	Not supported	Not supported	Not supported
Sinhala	General - Sinhala	16k	Supported	Not supported	Not supported	Not supported	Not supported
Catalan	General - Catalan	16k	Supported	Not supported	Not supported	Not supported	Not supported
Khmer	General - Khmer	16k	Supported	Not supported	Not supported	Not supported	Not supported
Hebrew	General - Hebrew	16k	Supported	Not supported	Not supported	Not supported	Not supported
Croatian	General - Croatian	16k	Supported	Not supported	Not supported	Not supported	Not supported
Hausa	General - Hausa	16k	Supported	Not supported	Not supported	Not supported	Not supported
Marathi	General - Marathi	16k	Supported	Not supported	Not supported	Not supported	Not supported
Telugu	General - Telugu	16k	Supported	Not supported	Not supported	Not supported	Not supported
Punjabi	General - Punjabi	16k	Supported	Not supported	Not supported	Not supported	Not supported
Swedish	General - Swedish	16k	Supported	Not supported	Not supported	Not supported	Not supported
Bulgarian	General - Bulgarian	16k	Supported	Not supported	Not supported	Not supported	Not supported
Danish	General - Danish	16k	Supported	Not supported	Not supported	Not supported	Not supported
Norwegian	General - Norwegian	16k	Supported	Not supported	Not supported	Not supported	Not supported
Kannada	General - Kannada	16k	Supported	Not supported	Not supported	Not supported	Not supported
Malayalam	General - Malayalam	16k	Supported	Not supported	Not supported	Not supported	Not supported
Czech	General - Czech	16k	Supported	Not supported	Not supported	Not supported	Not supported
Urdu	General - Urdu	16k	Supported	Not supported	Not supported	Not supported	Not supported
Nepali	General - Nepali	16k	Supported	Not supported	Not supported	Not supported	Not supported
Mongolian (Mongolia)	General - Mongolian (Mongolia)	16k	Supported	Not supported	Not supported	Not supported	Not supported
Uzbek	General - Uzbek	16k	Supported	Not supported	Not supported	Not supported	Not supported

Dialects

Language	Model name	Sample rate	Punctuation	ITN	Disfluency removal	Semantic sentence detection	Audio-text alignment
Cantonese	General - Cantonese	16k	Supported	Supported	Supported	Not supported	Supported
	Call center (general)	8k	Supported	Supported	Supported	Not supported	Supported
	Cantonese-Mandarin free speech	8k	Supported	Supported	Supported	Not supported	Not supported
	Southeast Asian languages	16k	Supported	Not supported	Not supported	Not supported	Not supported
Cantonese (Traditional Chinese)	General - Cantonese (Traditional Chinese)	8k	Supported	Not supported	Not supported	Not supported	Not supported
Cantonese (Traditional Chinese)	General - Cantonese (Traditional Chinese)	16k	Supported	Not supported	Not supported	Not supported	Not supported
Sichuan dialect	General - Sichuan dialect	16k	Supported	Supported	Supported	Supported	Supported
Sichuan dialect	Call center (general)	8k	Supported	Supported	Supported	Supported	Supported
Hubei dialect	General - Hubei dialect	16k	Supported	Supported	Supported	Supported	Supported
Hubei dialect	General - Hubei dialect	8k	Supported	Supported	Supported	Supported	Supported
Shanghainese	General - Shanghainese	16k	Supported	Supported	Supported	Supported	Not supported
Hunan dialect	General - Hunan dialect	16k	Supported	Supported	Supported	Supported	Supported
Henan dialect	General - Henan dialect	16k	Supported	Supported	Supported	Supported	Supported
Henan dialect	General - Henan dialect	8k	Supported	Supported	Supported	Supported	Supported
Zhejiang dialect	General - Zhejiang dialect	16k	Supported	Supported	Supported	Supported	Not supported
Northeastern Mandarin	General - Northeastern Mandarin	16k	Supported	Supported	Supported	Supported	Supported
Shandong dialect	General - Shandong dialect	16k	Supported	Supported	Supported	Supported	Supported
Tianjin dialect	General - Tianjin dialect	16k	Supported	Supported	Supported	Supported	Supported
Shaanxi dialect	General - Shaanxi dialect	16k	Supported	Supported	Supported	Supported	Supported
Shanxi dialect	General - Shanxi dialect	16k	Supported	Supported	Supported	Supported	Supported
Guizhou dialect	General - Guizhou dialect	16k	Supported	Supported	Supported	Supported	Supported
Yunnan dialect	General - Yunnan dialect	16k	Supported	Supported	Supported	Supported	Supported
Gansu dialect	General - Gansu dialect	16k	Supported	Supported	Supported	Supported	Supported
Suzhou dialect	General - Suzhou dialect	16k	Supported	Supported	Supported	Supported	Not supported
Hokkien	General - Hokkien	16k	Supported	Supported	Supported	Supported	Not supported
Jiangxi dialect	General - Jiangxi dialect	16k	Supported	Supported	Supported	Supported	Supported
Ningxia dialect	General - Ningxia dialect	16k	Supported	Supported	Supported	Supported	Supported
Guangxi dialect	General - Guangxi dialect	16k	Supported	Supported	Supported	Supported	Supported
Guangxi dialect	General - Guangxi dialect	8k	Supported	Supported	Supported	Supported	Supported
Chinese Mandarin	Shiyinshi V1 - end-to-end model, education content analysis, medical content analysis, news and media content analysis, entertainment video content analysis, upgraded offline audio and video transcription, new retail recognition, mobility recognition, and automotive recognition	16k	Supported	Supported	Supported	Supported	Supported
	Chinese-English free speech	16k	Supported	Supported	Supported	Supported	Not supported
	Shiyinshi V1 - end-to-end model	8k	Supported	Supported	Supported	Supported	Supported
	Southeast Asian languages	16k	Supported	Not supported	Not supported	Not supported	Not supported

Legacy version compatibility

Callback results for tasks with version="2.0" use snake-case fields rather than the camel-case fields used by version 4.0. Applications that parse the legacy callback format must update their field parsing when upgrading to 4.0. The following example shows a legacy callback:

{
        "result": [{
                "begin_time": 340,
                "channel_id": 0,
                "emotion_value": 5.0,
                "end_time": 2365,
                "silence_duration": 0,
                "speech_rate": 177,
                "text": "北京的天气。"
        }],
        "task_id": "3f5d4c0c399511e98dc025f34473****",
        "status_code": 21050000,
        "status_text": "SUCCESS",
        "request_time": 1551164878830,
        "solve_time": 1551164879230,
        "biz_duration": 2956
}