The automated review feature is implemented based on a large amount of labeled data and deep learning algorithms. This feature reviews media files from multiple dimensions such as voice, text, and image. This feature can accurately identify illegal and non-compliant content such as pornographic, terrorist, and politically sensitive content in videos and the corresponding thumbnails and titles. This feature is widely used in short video review and media review scenarios. This topic describes how to create a custom review template and perform automated reviews.
Scenarios
UGC review
The automated review feature can be used to detect illegal content and quality issues in a large amount of user-generated content (UGC). This can reduce the workload of manual review.
Content review for on-demand videos
The automated review feature can be used to detect terrorist and politically sensitive content in on-demand videos based on the political news. This can significantly reduce illegal content.
Review types
Review type | Description | Category of review results |
Pornography | Detects whether media files contain pornographic or sexy content. | Normal, porn, and sexy |
Terrorist and politically sensitive content | Detects whether media files contain terrorist or politically sensitive content. | Normal, blood and gore, explosion, special outfit, special logo, weapon, politics, violence, crowd, parade, car crash, flag, and location |
Ad | Detects whether video and image files contain ads or illegal text. | Normal, politics, porn, abuse, terrorism, prohibited content, junk content, spam, illegal ads, QR code, mini program code, and other ads |
Undesirable content | Detects whether video and image files contain undesirable scenes, such as the black screen, black bars, dark images, picture-in-picture (PiP), smoking, or streaming during driving. | Normal, PiP, smoking, streaming during driving, and no image such as black screen or white screen |
Logo | Detects whether video and image files contain logos, such as controlled media logos or trademarks. | Normal, controlled logo, and trademark |
Audio anti-spam | Detects whether audio and video files contain junk audio information, such as ads, politically sensitive content, or terrorist content. | Normal, junk information, ads, politics, terrorism, abuse, porn, spam, violation, meaningless, and custom such as custom keywords |
Billing
You are charged for the total duration of reviewed videos. To view billing details, submit a ticket. For more information, see Contact us.
Usage notes
The automated review feature is available only in the China (Shanghai), China (Beijing), and Singapore regions.
Only Illicit Content Identification, Violence, and Undesirable Content Detection are available in the Singapore region.
The final review result of an audio or video file is determined by the results of automated review and manual review. If the results of automated review and manual review are different, the manual review result takes precedence over the automated review result.
Only audio and video files in the Normal state can be played on other platforms. You can play audio and video files in the Blocked or Reviewing state only by using the ApsaraVideo VOD console or IP addresses in review security groups described in Security IP address preview. However, unauthorized users can play the audio or video files if the playback URLs are leaked. You cannot call API operations or use SDKs to obtain the playback URLs of audio or video files that are in the Blocked or Reviewing state.
Prerequisites
ApsaraVideo VOD is activated. For more information, see Activate ApsaraVideo VOD.
The media files that you want to review are uploaded to the media asset library in ApsaraVideo VOD. For more information, see Overview.
How to use the automated review feature
Step 1: (Optional) Configure an automated review template
ApsaraVideo VOD provides a built-in automated review template as the default template. You can also create a custom automated review template based on your business requirements.
You cannot change the configurations of the built-in automated review template or delete the template. The following content describes the configurations:
Review Type: Violence and Illicit Content Identification
Review Scope: Video, Cover, and Name
Review Content: Image
You can create up to 20 custom automated review templates.
Step 2: (Optional) Configure event notifications
Before you submit an automated review job, we recommend that you configure event notifications. This allows you to obtain the status and review result of the automated review job from the returned callback message. For more information about how to configure event notifications, see Configure callbacks.
The following content describes the events related to automated review for which you can configure notifications:
Video Analysis Completed. You can call the SubmitAIMediaAuditJob operation to submit a review job only after the VideoAnalysisComplete event is triggered.
AI Processing Completed. The AIMediaAuditComplete event is triggered every time an AI processing job is complete.
Step 3: Submit an automated review job
You can use a workflow to automatically submit an automated review job for an audio or video file that is being uploaded. Alternatively, you can submit an automated review job after the file is uploaded.
Step 4: View the review results
The image resources in the automated review results are stored in ApsaraVideo VOD for two weeks free of charge. Image resources that are stored for more than two weeks are automatically deleted.
If you have configured event notifications, you can obtain the automated review result from the AIMediaAuditComplete callback.