Video Moderation 2.0 and billing-AI Guardrails(AI Guardrails)-阿里云帮助中心

Video Moderation Premium Edition

Features

Video Moderation Enhanced detects content or elements in on-demand videos and live streams that violate regulations, disrupt platform content standards, or harm user experience. The service reuses configurations from Image Moderation Enhanced and Voice Moderation Enhanced. Through Video Moderation Enhanced, you can use the risk labels and confidence scores from the API to take moderation actions on video images and voice based on your industry standards or platform rules.

Version comparison

Compared to version 1.0, Video Moderation Premium Edition offers more risk categories, a wider range of risk labels, and more flexible configuration options in the console.

Item	Video Moderation Premium Edition	Video Moderation 1.0
Default number of concurrent streams	50	20
Default video size limit	500 MB	200 MB
Risk detection scope	Video images: Recommended services: Video/Live Stream Snapshot Detection Hybrid large-small model image moderation service Large model moderation: General large model image moderation service Hybrid large-small model image moderation service Hybrid large-small model image moderation service (outside the China site) General use cases: Common baseline detection Common baseline detection (Professional Edition) Common baseline detection (outside the China site) Content administration detection Business-specific use cases: Video/Live Stream Snapshot Detection Profile picture detection Post comment image detection Marketing material detection Special use cases: AI-generated image detection AI-generated image detection (video snapshot version) AI-generated image detection (Professional Edition) AI-generated image detection (Flagship Edition) Universal image recognition AIGC use cases (for AIGC-related risk detection, we recommend that you use AI Guardrails): AIGC image risk detection AIGC image copyright infringement detection Note Video Moderation Premium Edition integrates Image Moderation Premium Edition. For details about Image Moderation Premium Edition, see Service description. Video voice: Audio and video media detection Audio and video media multi-language detection Audio and video media detection (large model version) Social entertainment live stream detection Social entertainment live stream multi-language detection Social entertainment live stream detection (large model version) AI-generated voice detection Note Video Moderation Premium Edition integrates Voice Moderation Premium Edition. For details about Voice Moderation Premium Edition, see Service description.	Video images: Pornographic content detection Terrorism and political content detection Undesirable scene detection Logo detection Text and image violation detection Video voice: Voice violation detection
Console features	Configure video image detection services. Configure video voice detection services. Configure video snapshot settings. Configure result return settings.	Configure detection item settings.
Billing	Billing is based on video images + video voice (whether to enable video voice moderation is configurable). Total cost = (Number of captured frames x Number of services) + (Video duration x Unit price for video voice) Image moderation is billed separately for each service. The services are optional, and their prices are the same as those for Image Moderation Premium Edition. Video/Live Stream Snapshot Detection Common baseline detection Common baseline detection (Professional Edition) Common baseline detection (outside the China site) Content administration detection AIGC image risk detection AIGC image copyright infringement detection AI-generated image detection AI-generated image detection (video snapshot version) AI-generated image detection (Professional Edition) AI-generated image detection (Flagship Edition) Profile picture detection Post comment image detection Marketing material detection Universal image recognition General image moderation large model service Hybrid large-small model image moderation service Hybrid large-small model image moderation service (outside the China site) Video voice moderation is billed based on video duration. The price is 10% lower than the price of Voice Moderation Premium Edition.	Billing is based on image risk scenarios (scene) + video voice (whether to enable video voice moderation is configurable). Total cost = (Number of captured frames x Number of risk scenarios) + (Video duration x Unit price for voice moderation) Image moderation is billed separately for each risk scenario. The scenarios are optional, and their prices are 1.8 times the prices of the Image Moderation 1.0 service. Pornographic content detection Terrorism and political content detection Undesirable scene detection Logo detection Text and image violation detection Video voice moderation is billed based on video duration. The price is the same as that of Voice Moderation 1.0.

Service description

Video Moderation Premium Edition provides the following services.

Service	Detection content	Use cases
Video file detection (videoDetection)	Scans video files for violative visual or audio content.	Detect prohibited or inappropriate content in video files. We recommend that you perform this detection on all video files that are publicly accessible over the Internet.
Video file detection (outside the China site) (videoDetection_cb)	Scans video files for violative visual or audio content outside the Chinese mainland. Note This service is available in the Singapore (ap-southeast-1) and US (Virginia) (us-east-1) regions.	Detect prohibited or inappropriate content in video files. We recommend that you perform this detection on all video files that are publicly accessible outside the Chinese mainland.
Video file detection (large model version) (videoDetectionByVL)	Scans video files for violative visual or audio content by using the large model-powered image moderation service. Note This service is available in the Shanghai, Beijing, and Shenzhen regions.	Detect prohibited or inappropriate content in video files. You can configure moderation rules for the image moderation large model and the voice moderation large model.
Live stream moderation (liveStreamDetection)	Scans live video streams for violative visual or audio content.	Detect prohibited or inappropriate content in live video streams. We recommend that you perform this detection on all live streams that are publicly accessible over the Internet.
Live stream moderation (outside the China site) (liveStreamDetection_cb)	Scans live video streams for violative visual or audio content outside the Chinese mainland.	Detect prohibited or inappropriate content in live video streams. We recommend that you perform this detection on all live streams that are publicly accessible outside the Chinese mainland.
Live stream moderation (large model version) (liveStreamDetectionByVL)	Scans live video streams for violative visual or audio content by using the large model-powered image moderation service. Note This service is available in the Shanghai, Beijing, and Shenzhen regions.	Detect prohibited or inappropriate content in live video streams. You can configure moderation rules for the image moderation large model and the voice moderation large model.
Video file detection (large model, outside the China site) (videoDetectionByVL_cb)	Scans video files for violative visual or audio content by using the large model-powered image moderation service outside the Chinese mainland. Note This service is available in the Singapore (ap-southeast-1) and US (Virginia) (us-east-1) regions.	Detect prohibited or inappropriate content in publicly accessible video files outside the Chinese mainland. You can configure moderation rules for the large image model.
Live stream moderation (large model, outside the China site) (liveStreamDetectionByVL_cb)	Scans live video streams for violative visual or audio content by using the large model-powered image moderation service outside the Chinese mainland. Note This service is available in the Singapore (ap-southeast-1) and US (Virginia) (us-east-1) regions.	Detect prohibited or inappropriate content in publicly accessible live video streams outside the Chinese mainland. You can configure moderation rules for the image moderation large model and the voice moderation large model.
AI-generated video detection (videoAigcDetector)	Analyzes input videos to determine if they are likely generated by AIGC or contain an AI-generated implicit watermark. If an implicit watermark is detected, the service returns its metadata. Note This service is available in the Beijing and Shanghai regions.	Determine whether a video is likely AIGC-generated. We recommend using this service when you need to detect and label AI-generated synthetic content in videos.

Billing

Pay-as-you-go

After you activate Video Moderation Premium Edition, services are billed on a pay-as-you-go basis by default. You are charged only for the API calls you make. The following table describes the pay-as-you-go pricing for the Video Moderation Premium Edition API.

Moderation type	Supported services	Price
Standard video image detection (image_standard, video_image_standard)	Common baseline detection: baselineCheck Content administration detection: tonalityImprove AIGC image detection: aigcCheck AIGC image copyright infringement detection: aigcViolationDetection AI-generated image detection: aigcDetector	CNY 15.00/10,000 calls Note Each call to a listed service counts as one moderation instance. For example, 100 calls to the common baseline detection service cost CNY 0.15.
Premium video image detection (image_advanced, video_image_advanced)	Video/Live Stream Snapshot Detection: liveStreamCheck Common baseline detection (Professional Edition): baselineCheck_pro Common baseline detection (outside the China site): baselineCheck_cb Profile picture detection: profilePhotoCheck Marketing material detection: advertisingCheck Post comment image detection: postImageCheck Universal image recognition: generalRecognition AI-generated image detection (video snapshot version): aigcDetectorForFrame	CNY 30.00/10,000 calls Note Each call to a listed service counts as one moderation instance. For example, 100 calls to the common baseline detection (Professional Edition) service cost CNY 0.30.
Video image moderation large model version (image_vl_standard)	Hybrid large-small model image moderation service: postImageCheckByVL Hybrid large-small model image moderation service (outside the China site): postImageCheckByVL_cb General image moderation large model service: baselineCheckByVL	CNY 45.00/10,000 calls Note Each call to a listed service counts as one moderation instance. For example, 100 calls to the hybrid large-small model image moderation service cost CNY 0.45.
Standard video voice moderation (video_standard)	Audio and video media detection: audio_media_detection Live stream detection: live_stream_detection	CNY 202.50/10,000 minutes, equivalent to CNY 1.215/hour.
Premium video voice moderation (video_advanced)	Audio and video media multi-language detection: audio_multilingual_cb Live stream multi-language detection: stream_multilingual_cb AI-generated voice detection: voice_aigc_detector	CNY 337.50/10,000 minutes, equivalent to CNY 2.025/hour.

Note

For the pay-as-you-go billing method of Content Moderation Premium Edition, usage is metered and billed every 1 hours. On your bill details page, the moderationType field corresponds to the moderation types listed in the preceding table. For more information, see Bill Details.

Billing calculation logic

Video Moderation Premium Edition bills video image moderation and video voice moderation independently:

Video image cost = Number of captured frames × Number of services × Unit price
Video voice cost = Video duration × Unit price (the video voice cost is not multiplied by the number of services)

In billing, "calls" refers to the number of frames extracted from the video at the configured frame capture rate. The default frame capture rate is 1 frame per second (adjustable in the console). Total number of captured frames = Video duration (seconds) × Frame capture rate (frames/second).

Billing example: For a 40-second video with the default frame capture rate (1 frame per second) and 1 image moderation service enabled:

Billing item	Formula	Example calculation
Video image cost	Number of captured frames × Number of services × Unit price	40 frames × 1 (number of services) × unit price of the service
Video voice cost	Video duration (minutes) × Unit price	40 seconds (billed as 1 minute if less than 1 minute) × voice unit price

To verify the cost composition of a single moderation task, filter the call records by TaskId in Bill Details.

Note

Video Moderation Premium Edition uses exactly the same billing standards across all regions (including the Chinese mainland, Singapore, and the United States). All regions are billed using the same formula: (Number of captured frames × Number of services × Unit price) + (Video duration × Voice moderation unit price).

Resource plans

If you have a large or predictable volume of video moderation tasks, we recommend that you purchase a resource plan in advance. Larger resource plans offer greater discounts, and the unit price of a resource plan is lower than that of pay-as-you-go billing. For more information, see purchase Content Moderation Premium Edition resource plans.

Content Moderation Premium Edition resource plans offset usage of Premium Edition services. They cannot be shared with Content Moderation 1.0 resource plans. The following table lists the offset factors.

Moderation type	Offset factor
Standard video image moderation (image_standard, video_image_standard)	2: Each successful call to any of the services listed on the left deducts 2 calls from the resource plan capacity. For example, if the resource plan capacity is 100 calls, one successful call to Common baseline detection deducts 2 calls, leaving 98.
Premium video image moderation (image_advanced, video_image_advanced)	4: Each successful call to any of the services listed on the left deducts 4 calls from the resource plan capacity. For example, if the resource plan capacity is 100 calls, one successful call to Video/Live Stream Snapshot Detection deducts 4 calls, leaving 96.
Video image moderation large model version (image_vl_standard)	6: Each successful API call deducts 6 calls from the resource plan capacity. For example, if the resource plan capacity is 10 calls, one successful API call deducts 6 calls, leaving 4.
Standard video voice moderation (video_standard)	27: Each minute of successfully processed content in the services listed on the left deducts 27 calls from the resource plan capacity. For example, if the resource plan capacity is 100 calls, successfully moderating 1 minute of video voice deducts 27 calls, leaving 73.
Premium video voice moderation (video_advanced)	45: Each minute of successfully processed content in the services listed on the left deducts 45 calls from the resource plan capacity. For example, if the resource plan capacity is 100 calls, successfully moderating 1 minute of video voice deducts 45 calls, leaving 55.

After you purchase a resource plan, it is used first to cover fees for your API calls. If your resource plan is depleted, the system automatically switches to pay-as-you-go billing. Monitor your resource plan balance and pay-as-you-go bills. You can configure low-balance alerts in the Resource Plan System.

FAQ

Why did I incur unexpectedly high charges after publishing a single video?

A: High charges are usually caused by the following reasons:

Common causes

Repeated API calls: Changes in your business code logic may cause the same video to be submitted for moderation multiple times. Each call is billed independently.
High frame capture rate: A short frame capture interval (for example, the default 1 second per frame) generates a large number of captured frames for long videos, significantly increasing image moderation costs. For example, a 1-hour video captured at 1 frame per second produces 3,600 frames, and the cost increases linearly with the video duration.

Troubleshooting suggestions

Check whether your business code contains logic that repeatedly submits the same video for moderation to avoid duplicate charges.
Verify that the frame capture interval in the console or in the API request parameters meets your expectations (a longer interval means fewer captured frames and lower costs, but also increases the risk of missed violations).
Check the call volume and cost sources of specific moderation requests by TaskId in Bill Details.

Usage

Integration

You can integrate Video Moderation Premium Edition in the following ways:

Use an SDK. For more information, see Video Moderation Premium Edition 2.0 SDK and integration guide.
Call the API directly. For more information, see Video Moderation Premium Edition 2.0 API reference.

Console operations

When using the service for the first time, configure your moderation settings in the User Console.
You can modify moderation policies, configure different policies for various business needs, view results, and track usage.

For detailed instructions, see Console Operation Guide.