Video Moderation 2.0 ships with a pre-configured general policy built on industry-wide content standards. When the default policy does not fit your business requirements, customize it through the console. This guide covers how to adjust detection policies, split rules across multiple business lines, query moderation results, and monitor usage statistics.
Prerequisites
Before you begin, ensure that you have:
An active Video Moderation 2.0 service. To activate it, go to the Content Moderation 2.0 page.
Familiarity with the billing model. See Introduction and billing for Video Moderation Enhanced Edition 2.0 before activating.
Use cases
| Scenario | What you can do |
|---|---|
| Adjust threat detection policies | Change image snapshot frequency, enable or disable voice detection, and choose which results to return. |
| Separate policies for multiple business lines | Copy the default service to create independent configurations for each business operation. |
| Query moderation results | Search and filter recent moderation results by task ID, data ID, service, label, or threat level. |
| Monitor usage statistics | View and export usage data by day or month, and track threat-level trends across your video content. |
Adjust threat detection policies
Each video moderation service has configurable rules for image detection, voice detection, and result output. Adjust these to match your business requirements.
Log on to the AI Guardrails console.
In the navigation pane, choose Machine Moderation V2.0 > Video Moderation > Rules.
Find the video file moderation (videoDetection_global) service and click Settings in the Actions column.videoDetection_globalvideoDetection_globalvideoDetection_global_01videoDetection_global_01videoDetection_globalvideoDetection_globalvideoDetection_globalvideoDetection_global_02videoDetection_global_01videoDetection_globalvideoDetection_global
On the rules page, adjust the configurations as needed. Example: To take 1,000 snapshots per video, scan frames for serious violations, scan audio for violations, and return all results: For details on image moderation rules, see Introduction to Image Moderation 2.0 and its billing.
For Image Snapshot Method, select Fixed Total Number of Snapshots, enter 1000, and set Image Moderation Rule to General Baseline Detection.
Turn on the Moderate Audio switch and select Audio and Video Media Moderation.
Select Return All Results.
Click Save.
The new rule takes effect within 2 to 5 minutes.
General Baseline Detection targets serious violations and is suited for businesses that prioritize precision — flagging only clear-cut content reduces manual review overhead. Return All Results gives you complete visibility into every moderation decision, including content that passed. Use Return All Results when you need audit trails, want to analyze borderline content, or prefer higher recall at the cost of more results to review.
Set different policies for multiple business lines
Copy the default service to create independent configurations for each business operation. Each copy runs its own detection rules.
Log on to the Content Moderation consoleContent Moderation consoleContent Moderation console.
In the navigation pane, choose Machine Moderation V2.0 > Video Moderation > Rules.
Find the videoDetection_global service and click Copy in the Actions column.
2.0In the Copy Service dialog box, enter a Service Name and Service Description, then click OK.
Once created, find the new service (for example, videoDetection_global_01), click Settings, and adjust its rules as needed.
2.0
The copied service is callable 1 to 2 minutes after creation. Repeat this process to create as many independent service configurations as your business requires.
Example: If business operations A, B, and C all use video moderation but need different detection standards, copy videoDetection_global twice to create videoDetection_global_01 and videoDetection_global_02, then configure each with its own rules.
Query moderation results
Search moderation results to investigate specific videos or identify content patterns.
Log on to the Content Moderation consoleContent Moderation consoleContent Moderation console.
In the navigation pane, choose Machine Moderation V2.0 > Video Moderation > Detection Results.
On the Detection Results page, set query conditions and search. Filter by:
taskId,dataId, service, image label, voice label, or video threat. Example: To find all results with frame-level hits, set the query condition to Image Label is not equal to nonLabel. Separate multiple labels with commas (,).In the results list, do one of the following: The Screen review results tab shows each frame snapshot and the labels that were hit. The Voice Audit Results tab shows speech-to-text transcripts and the labels that were hit.
Click Details in the Actions column to view high-level moderation information for a video.
Click Audio and Image Results to see frame-level and audio-level detail.
Results are sorted in reverse chronological order. A maximum of 50,000 records are shown, covering the last 30 days. For longer-term analysis, store the data or logs from each API call in your own data pipeline.
Voice moderation results appear only if voice moderation is enabled in the service settings.AI Guardrails console
View usage statistics
The Dashboards page provides data to help you track call volume, evaluate whether your detection policy is too strict or too lenient, and inform decisions about policy adjustments. Use it to monitor block rates (threat statistics), service call volume (curve chart), and content distribution by threat type (threat distribution).
videoDetection_globalLog on to the Content Moderation consoleContent Moderation consoleContent Moderation console.
In the navigation pane, choose Machine Moderation V2.0 > Video Moderation > Dashboards.
Select a time range. The page shows the following: Query usage View usage by day or month, filtered by Alibaba Cloud account, RAM user, or service. Statistical data is stored for one year; each query covers up to two months.
The curve chart shows the number of video moderation tasks (videos submitted).
The column chart shows image and voice moderation usage — number of video snapshots scanned and audio duration moderated (in minutes).
Export usage Click
to export usage data by day or month. The exported Excel report includes only periods with recorded usage. View threat statistics Use this section to identify which threat categories are most prevalent and how threat levels shift over time. This helps you evaluate whether your current detection policy needs adjustment.Threat Level Distribution: Shows the count and percentage of high, medium, low, and no-threat results for the selected time range.
Threat Trend Statistics: Shows threat counts by day or month, so you can spot spikes or declining trends.
Threat Distribution Statistics: Shows the top 5 threat types and their share of all detected threats, across both video frames and audio.
Field Description Unit Account UID UID of the account for which data is exported None service Video moderation service / Video image moderation service / Video voice moderation service None Usage Total number of calls Count Date Date of the statistics Day/Month