ApsaraVideo VOD:Media fingerprint

Features

A media fingerprint is typically a binary string that uniquely identifies a video. The media fingerprint service extracts and compares fingerprint features from the images and audio in videos. This process helps solve issues such as finding duplicate videos, tracing the source of video clips, and identifying original content.

• Uniqueness: Different videos have different media fingerprints.

• Stability: The fingerprint remains unchanged despite transformations such as format conversion, editing, splicing, compression, or rotation of the audio or video file.

Scenarios

• Duplicate video detection

  You can build an independent media fingerprint library and perform rapid comparisons of sample videos within it to recall duplicate videos or segments in real time. This prevents an excessive number of identical or similar videos during content delivery and personalized recommendation, which helps maintain a positive user experience.

• Original content identification

  You can accurately identify whether a video is original, has been edited or transformed, or is a re-creation by we media. This capability addresses copyright enforcement challenges in the short video industry, provides a technical foundation for the video asset trading ecosystem, and protects the interests of original creators and video platforms against infringing content.

• Quick moderation

  You can maintain a library of non-compliant videos or a custom blacklist. You can then quickly compare new videos against this library using DNA matching to detect violations. Compared with traditional moderation, this approach improves efficiency when you review large volumes of videos, ensures accuracy, and reduces moderation costs.

• Ad revenue sharing

  You can use media fingerprinting to search an ad revenue-sharing library and monitor or identify specific advertisements. This supports a dynamic ad revenue-sharing business ecosystem, lets you easily control ad delivery timing and frequency, and protects the interests of both ad copyright holders and advertising platforms. Additionally, it enables efficient ad identification, replacement, and ad slot auctions.

• Video source tracing

  For newly added videos, you can perform fast searches in a large-scale media fingerprint library to locate the video's propagation path. This effectively addresses challenges such as video source tracing, path analysis, and uncovering relationships among media resources.

Preparations

• Complete the steps described in the Notes section of the video AI overview. Then, go to the Media Fingerprint product page and click Request Activation - Use ApsaraVideo VOD Service. Fill in the required information, submit your request, and wait for a notification.

• When you use media fingerprinting for the first time, you must provide your user ID (UID) and region information. Submit a ticket to request free activation of the media fingerprint library. If you do not activate the library, your subsequent use of the media fingerprint feature will be affected.

Billing

Media fingerprinting is billed based on the duration of successfully processed videos. For more information about billing, see Video AI billing.

How to use

• Automatic trigger

  After you activate the media fingerprint service, any video that you upload to ApsaraVideo VOD automatically triggers media fingerprint processing after transcoding is complete. The result is delivered through a message callback. For more information about the callback message format, see Media fingerprint job complete.

• API trigger

  • For videos that are already uploaded to ApsaraVideo VOD, you can call the SubmitAIJob operation to submit a media fingerprint job.

  • After you submit the job, the media fingerprint job is processed asynchronously in the background. The final result is sent through a message callback. For more information about the callback message format, see Media fingerprint job complete. You can also call the ListAIJob operation to query the job result.

    Note

    • For automatic triggers, you must configure transcoding to output only one MP4 stream, such as a low definition or original quality stream. Media fingerprint processing starts after one stream finishes transcoding.

    • If you use the API method, you do not need to configure transcoding. You can process the source file directly.

• Query results

  • After a media fingerprint job is complete, you can call the GetMediaDNAResult operation to retrieve the DNA result.

  • The video fingerprinting result includes the following information:

    The media fingerprint processing result includes the following main information:

    • Whether the job matched any duplicate videos.

    • If a match exists, details of the duplicate video are returned.

    • The timecode intervals where duplication occurs between videos.

    • The degree of duplication between videos, expressed as a number from 0 to 1.

• Existing videos

  To ensure high-quality DNA comparison results, you can import existing videos to build a complete feature library. You can use one of the following two methods:

• If the total duration of your existing video content is less than 1 million minutes, you can directly call the SubmitAIJob operation to process the videos.

• If the total duration of your existing video content exceeds 1 million minutes, submit a ticket for technical support. In the ticket, you must specify the quantity, total duration, and storage location of your existing videos.