Skip to main content

Content Moderation

Content moderation, based on AI recognition technology, helps you automatically detect illegal content in files to ensure business security.

Supported Moderation Types​

Moderation TypeDescriptionDetection Content
Image ModerationDetect illegal content in imagesPornographic, political, violent, illegal, advertising, etc.
Video ModerationDetect illegal content in videosPornographic, political, violent, illegal, advertising, etc.
Audio ModerationDetect illegal content in audioPornographic, political, abusive, illegal, advertising, etc.
Text ModerationDetect sensitive words in textPornographic, political, abusive, illegal, advertising, etc.
Document ModerationDetect illegal content in documentsSupports PDF, Word, Excel, PPT, and other formats

Main Features​

  • Auto Recognition: Automatically identify illegal content based on deep learning technology
  • Custom Policies: Configure moderation policies and sensitive word libraries according to business needs
  • Real-time Moderation: Support automatic moderation upon upload, blocking illegal content immediately
  • Callback Notification: Real-time callback for moderation results, quick response
  • Auto Processing: Support automatic freezing or deletion of illegal content

Use Cases​

  • UGC Content Moderation: Moderate images, videos, and text content uploaded by users
  • Social Platforms: Security moderation of comments, posts, and chat content
  • E-commerce Platforms: Compliance moderation of product images and description information
  • Live Streaming Platforms: Real-time moderation of live content and bullet comments

Configuration Steps​

Console Access Path​

CloudDev Platform - Cloud Storage → Click "Go to COS Bucket" → Left Menu "Content Moderation" → "Auto Moderation Configuration" → Select Moderation Type (Image/Video/Audio/Text/Document)

📌 Note

All content moderation features require accessing the COS console first:

  1. In CloudDev Platform - Cloud Storage page, click "Go to COS Bucket"
  2. In the COS console left menu, select "Content Moderation" → "Auto Moderation Configuration"
  3. Select the corresponding moderation type (image moderation, video moderation, audio moderation, text moderation, document moderation)

Image Moderation Configuration​

1. Enter Image Moderation

  • In the left menu, select "Content Moderation" → "Auto Moderation Configuration" → "Image Moderation"

2. Create Moderation Task

  • Click the "Create Task" button
  • Select moderation type: incremental moderation (auto-moderate newly uploaded images) or stock moderation (moderate existing images)
  • Configure moderation policy: select scenes to moderate (pornographic, political, violent, illegal, advertising, etc.)

3. Set Moderation Result Processing

  • Configure callback address: receive moderation result notifications
  • Set auto-freeze rules: automatically freeze illegal content
  • Select moderation result storage location

4. Enable Task

  • After confirming the configuration is correct, click "OK"
  • Task takes effect automatically after creation

API Call Method:

Video Moderation Configuration​

1. Enter Video Moderation

  • In the left menu, select "Content Moderation" → "Auto Moderation Configuration" → "Video Moderation"

2. Create Moderation Task

  • Click "Create Task"
  • Set moderation scope: specify file path or select entire bucket
  • Configure moderation policy: pornographic, political, violent, illegal, advertising, etc.

3. Configure Moderation Parameters

  • Frame Capture Method: Set video frame capture interval (e.g., capture one frame every 5 seconds)
  • Moderation Dimensions: Select frame moderation, audio moderation, or both
  • Callback Settings: Configure moderation result callback URL

4. Enable and Monitor

  • Task executes automatically after saving configuration
  • View moderation progress and results in task list

API Call Method:

Audio Moderation Configuration​

1. Enter Audio Moderation

  • In the left menu, select "Content Moderation" → "Auto Moderation Configuration" → "Audio Moderation"

2. Create Moderation Task

  • Click "Create Task"
  • Select audio files or folders to moderate

3. Configure Moderation Policy

  • Select moderation scenes: pornographic, political, abusive, illegal, advertising, etc.
  • Set voice recognition language (Chinese, English, etc.)
  • Configure sensitive word library

4. Set Callback and Processing Rules

  • Configure moderation result callback address
  • Set auto-processing rules for illegal content
  • Save and enable task

API Call Method:

Text Moderation Configuration​

1. Enter Text Moderation

  • In the left menu, select "Content Moderation" → "Auto Moderation Configuration" → "Text Moderation"

2. Configure Moderation Policy

  • Click "Moderation Configuration"
  • Select moderation type: pornographic, political, abusive, illegal, advertising, etc.

3. Custom Sensitive Word Library

  • Add business-related sensitive words
  • Set word library priority
  • Configure hit rules

4. API Call Method

  • Text moderation is mainly called through API
  • Integrate text moderation API in application

API Call Method:

Document Moderation Configuration​

1. Enter Document Moderation

  • In the left menu, select "Content Moderation" → "Auto Moderation Configuration" → "Document Moderation"

2. Create Moderation Task

  • Click "Create Task"
  • Select document files to moderate

3. Configure Moderation Parameters

  • Supported formats: PDF, Word, Excel, PPT, etc.
  • Select moderation scene and policy
  • Set moderation result notification method

4. View Moderation Results

  • View detailed moderation report after task completion
  • Support viewing illegal content by page
  • Export moderation result report

API Call Method:

Best Practices​

1. Combine Incremental and Stock Moderation​

  • Incremental Moderation: Configure auto-moderation for real-time detection of newly uploaded files
  • Stock Moderation: Regularly perform full scans of existing files

2. Custom Sensitive Word Library​

Add industry-related sensitive words according to business characteristics:

  • E-commerce platforms: words related to counterfeit goods, imitations
  • Social platforms: personal attacks, discriminatory words
  • Financial platforms: words related to illegal fundraising, fraud

3. Manual Review Mechanism​

For "suspected illegal" content:

  1. Receive moderation results through callback
  2. Put suspected illegal content into manual review queue
  3. Make final decision after manual review

4. Tiered Processing Strategy​

Take different measures based on the degree of violation:

  • Minor Violations: Warn users, request modification
  • Moderate Violations: Delete content directly, record violation count
  • Severe Violations: Freeze account, report to relevant departments

FAQ​

Q: How long does moderation take?​

A:

  • Image Moderation: Usually completed within 1-3 seconds
  • Video Moderation: Typically 10%-20% of video duration, based on video length
  • Audio Moderation: Typically 10%-20% of audio duration, based on audio length
  • Document Moderation: Typically 1-2 seconds per page, based on document page count

Q: How accurate is the moderation?​

A: Cloud Infinite is based on Tencent's years of accumulated massive data and deep learning technology:

  • Image Moderation Accuracy: ≥ 99%
  • Video Moderation Accuracy: ≥ 98%
  • Text Moderation Accuracy: ≥ 99%

Q: How to reduce moderation costs?​

A: Recommend the following strategies:

  1. Only moderate user-uploaded content, system content does not need moderation
  2. Adjust moderation strategy based on user credit level (high-credit users can reduce moderation frequency)
  3. Use custom policies to only moderate necessary scenes

Q: Is private deployment supported?​

A: Cloud Infinite content moderation is a cloud service and does not currently support private deployment.