Add identifiers to AIGC audio-Secure Access Service Edge(SASE)-阿里云帮助中心

This document describes how to add data identifiers to audio files.

Background

According to the "Measures for Identifying AI-Generated and Synthesized Content", starting from September 1, 2025, all content generated by artificial intelligence (AI), such as images, audio, and videos, must include both visible and invisible identifiers. These identifiers can be metadata tags or digital watermarks. This ensures that the source of the content is traceable and identifiable.The Secure Access Service Edge (SASE) platform provides the CreateWmEmbedTask API. This API helps enterprises manage content identifiers efficiently and meet compliance requirements.

Audio watermark parameter descriptions

Name		Type	Required	Description	Example value
WmType		string	Yes	The watermark type. Valid values: PureDocument: Document watermark. PureImage: Image watermark. PureAudio: Audio watermark. PureVideo: Video watermark. AigcDocument: AIGC document watermark. AigcImage: AIGC image watermark. AigcAudio: AIGC audio watermark. AigcVideo: AIGC video watermark.	PureAudio
Filename		string	Yes	The name of the file to embed the watermark in. The backend uses the file extension to verify the file type.	abc.mp3
FileUrl		string	Yes	The URL from which the file can be downloaded. The URL must be accessible over the public network.	https://example.com/abc.mp3
WmInfoSize		long	No	The bit width of the watermark information. The default value is 32. This parameter must be the same for both embedding and extracting. For example, if you use a 40-bit software development kit (SDK) for embedding, you must also set this value to 40 for extracting. Enumeration values: 32 40 64	32
WmInfoUint		string	No	The watermark information in decimal format. If you set this parameter, you cannot set WmInfoBytesB64. The value range depends on the WmInfoSize parameter: If WmInfoSize is 32, the value range is 1 to 4294967295. If WmInfoSize is 40, the value range is 1 to 1099511627775. If WmInfoSize is 64, the value range is 1 to 18446744073709551615.	123***
WmInfoBytesB64		string	No	The watermark information as a Base64-encoded string. If you set this parameter, you cannot set WmInfoUint.	aGVsbG8gc2F*****
ImageEmbedLevel		long	No	Image watermark parameter: A larger value provides higher robustness but reduces visual quality. The default value is 2. Valid values: 0 to 4.	2
ImageEmbedJpegQuality		long	No	Image watermark parameter: The desired JPEG compression quality factor for the output image. The default value is 95. Valid values: 1 to 100.	95
AudioControl		object	No	Audio control parameters.
	MetadataControl	object	No	Audio metadata control parameters.
	XmpKvBase64	string	No	The metadata in Base64 format. For the format before conversion, see the following code: `AIGC = { "Label": "1", "ContentProducer": "AXXXX", "ProduceID": "BXXXX", "ReservedCode1": "CXXX", "ContentPropagator": "DXXX", "PropagateID": "EXXX", "ReservedCode2": "FXXXX" }` Note The "AIGC=" prefix is required. Without it, the metadata cannot be added. Note that this prefix is different from the one used for image metadata. Use standard Base64 format with padding.	QUlHQz17IkxhYm********ZC85WkNpSm9ZcXlrSG5vcEpnQllZWE1BPT0ifQ==
	Enable	boolean	No	Specifies whether to enable the feature. true: The item is displayed. false: The item is not displayed.	true

Metadata format description

The XmpKvBase64 field requires metadata in Base64 format. Before you convert the data to Base64, it must meet the following format requirements.

Key	Description	Example value
Label	Indicates whether the content is AI-generated: 1: The content is generated and synthesized by artificial intelligence. 2: (For distribution platforms only) The content may be generated and synthesized by artificial intelligence. 3: (For distribution platforms only) The content is suspected to be generated and synthesized by artificial intelligence.	1
ContentProducer	The code or name of the service provider, used to identify the content producer. Required fields Digits 1-2: The identifier format definition code, which records the encoding rule version. For codes created using this guide, this is fixed to "00". Digit 3: The entity type. "1" for an organization, "2" for an individual. Digit 4: Set to "1" when the organization is identified by its Unified Social Credit Code. Digits 5-22: The organization's Unified Social Credit Code. Extension fields Digit 23: "1" for content generation and synthesis services, "2" for content distribution. Digits 24-27: The service or model number (must be unique). For more information, see Cybersecurity Standards Practice Guide.	001191******M000100Y43
ProduceID	The content production ID. A unique identifier used by the production platform to trace the synthesized content.	123456
ReserveCode1	Reserved field. Can store information for the generation and synthesis service provider to implement security protection and ensure content and identifier integrity. A hashing mechanism based on ContentProducer and ProduceID can be used for secure storage and verification of key information.	d41d8cd98f00b204e9800998ecf8427e
ContentPropagator	The name, number, or code of the distribution platform. For services that provide AI-generated synthetic content, this can be the same as ContentProducer.	001191******M000100Y43
PropagatorID	The content distribution ID. A unique number assigned by the distribution platform to the distributed synthetic content.	123456
ReserveCode2	Reserved field. Can be used by the content distribution service provider to implement security protection and ensure content and identifier integrity. A hashing mechanism based on ContentProducer and ProduceID can be used for secure storage and verification of key information.	d41d8cd98f00b204e9800998ecf8427e

For example, consider a production platform company named AA with the Unified Social Credit Code 91******M000100Y43.

AIGC={
	"Label": "1",
	"ContentProducer": "0011913301****",
	"ProduceID": "LGg1FYm0Tj-s4Ek****",
	"ReservedCode1": "LBkc9peJ0GoxPECtYBct7vfZZgZvwy4FZrhe/DJovt****",
	"ContentPropagator": "001191330106MA2C****",
	"PropagateID": "LGg1FYm0Tj-s4Ek_****",
	"ReservedCode2": "LBkc9peJ0GoxPECtYBct7vfZZgZvYYXMA****"
}

Example

{
  "WmType": "PureAudio",
  "FileUrl": "https://******.oss-cn-hangzhou.aliyuncs.com/test.mp3",
  "Filename": "test.mp3",
  "WmInfoSize": 32,
  "WmInfoUint": "1234567890",
  "InvisibleEnable": true,
  "AudioControl": {
    "MetadataControl": {
      "XmpKvBase64": "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",
      "Enable": true
    }
  }
}

Verification

Because you cannot add visible watermarks to audio files, you must use the AI-Generated Synthetic Content Identifier Service Platform to verify the audio file with the embedded identifier. If the identifier information is successfully extracted, the audio identifier embedding task is complete and valid.

After the watermark embedding task is successful, copy the value of the TaskId parameter from the response.
Call the GetWmEmbedTask - Query watermark embedding task details operation. Use the TaskId value as a parameter to retrieve the URL of the watermarked file (FileUrl).
Open the FileUrl in a browser to download the audio file.
Log on to the AI-Generated Synthetic Content Identifier Service Platform.
In the top navigation bar, choose Identifier Task > Audio Identifier Task.
Upload the audio file and click Start Detection to view the results.