This topic covers Manual Review 2.0, its human-machine review feature, and billing information.
Manual review 2.0
Overview
Manual Review 2.0 helps you identify content in images, text, audio, and videos that violates applicable online content policies, disrupts platform order, or degrades user experience. Compared with automated review, manual review provides higher accuracy and recall rates, but it takes longer to complete. Manual Review 2.0 from Content Moderation helps you mitigate business risks based on the manual review results.
Manual review: Call the Manual Review 2.0 API to submit content for review. The API returns the manual review results.
Human-machine review: Configure the human-machine review feature in the automated content moderation service. The system automatically forwards results from automated API reviews that meet specific conditions for manual review.
Before you use this service, contact your account manager to confirm the review standards and scope of content.
Available services
Manual Review 2.0 supports the following services:
Service |
Detection content |
Scenarios |
Manual image review (imageManualCheck) |
Detects pornographic, sexually suggestive, political, violent, terrorist, prohibited, abusive, and advertising content in images. |
Ideal for image moderation scenarios that require higher accuracy and recall rates. |
Manual long text review (longTextManualCheck) |
Detects pornographic, sexually suggestive, political, violent, terrorist, prohibited, abusive, and advertising content in text. |
Ideal for long text moderation scenarios that require higher accuracy and recall rates. |
Manual audio file review (voiceFileManualCheck) |
Detects pornographic, sexually suggestive, political, violent, terrorist, prohibited, abusive, and advertising content in audio. |
Ideal for audio file moderation scenarios that require higher accuracy and recall rates. |
Manual video file review (videoFileManualCheck) |
Detects pornographic, sexually suggestive, political, violent, terrorist, prohibited, abusive, and advertising content in video. |
Ideal for video file moderation scenarios that require higher accuracy and recall rates. |
Human-machine review supports all image, text, audio, and video moderation services. You can configure the human-machine review feature on the human review configuration tab of the relevant service. For more information, see Configure human review.
Billing
After you enable Manual Review 2.0, the default billing method is pay-as-you-go. Charges apply only when you use the service. The Manual Review 2.0 API is billed as follows.
Review type |
Business scenarios |
Unit price |
Manual image review_t0 (image_manual_annotation_t0) |
imageManualCheck: Manual image review |
CNY 0.1 per review |
Manual text review_t0 (text_manual_annotation_t0) |
longTextManualCheck: Manual long text review |
CNY 0.1 per review |
Manual audio review_t0 (voice_manual_annotation_t0) |
voiceFileManualCheck: Manual audio file review |
CNY 0.833 per minute |
Manual video review_t0 (video_manual_annotation_t0) |
videoFileManualCheck: Manual video file review |
CNY 1.333 per minute |
Pay-as-you-go bills for Manual Review 2.0 are generated every hour. In your billing details, the moderationType field corresponds to the review type field in the preceding table. For more information, see Billing Details.
Usage
Integration methods
Manual Review 2.0 supports the following integration methods:
Integrate the service using SDKs. For more information, see Manual review 2.0 SDK and access guide.
Integrate the service using APIs. For more information, see Manual review 2.0 asynchronous detection API.
Configure human-machine review workflows to integrate the manual review service. For more information, see Configure human review.