tencent cloud

Audio Moderation System
Audio Moderation System (AMS) can automatically recognize content in audio that may be offensive, unsafe, or inappropriate. It allows you to customize blocked keywords to recognize custom types of audio content.
Get StartedDocumentation
Features
Scenarios
Help and Documentation
Pricing
FAQs
Features
Audio Moderation System
Pornographic Recognition

Recognizes obscene words in audio. This greatly reduces your content operations costs and is suitable for various on-demand/live video and audio use cases.

Audio Moderation System
Ad Detection

Identifies traffic-diverting, spam, phishing, and fraud ads in audio through audio transcription to maintain a safe user experience.

Audio Moderation System
Vulgarity Detection

Recognizes filthy, vulgar, and other sexually suggestive content contained in audio.

Scenarios

Recognizes the speeches of hosts to identify the information that is not suitable for dissemination in real time, so as to protect the security of platform content operations.

Audio Moderation System

Recognizes speeches in game chat channels to block harmful content such as phishing, traffic-diverting, and pornographic ads.

Audio Moderation System
Pricing

AMS is billed in a pay-as-you-go manner. For more details, see Purchase Guide.

FAQs
What should I do if the error "Package delivery failed" is displayed?

Two possible reasons:

  • Your account balance
  • Your account balance is zero (it cannot be negative).

Solution: Top up your account.

What languages does AMS support?

AMS supports English, Chinese, Indonesian, Hindi, Bengali, Arabic, Japanese, Russian, Thai and Vietnamese.

Does AMS support customizing sensitive word dictionaries?

Yes. Go to “Custom Library Management” > “Custom Dictionary” in the AMS console, create a custom dictionary, and add keywords. After creating the custom dictionary, you need to create a custom moderation policy and bind the dictionary to the policy for the dictionary to take effect.

Is manual recognition supported? How do I access it?

Yes, and a separate manual recognition API needs to be accessed. You can submit a ticket to get the access documentation.

Is manual recognition a paid service?

Yes. You can contact us for pricing information.

Can I configure callbacks for manual recognition based on different push domain names?

Yes. You can create multiple sub-accounts under the root account and use them to configure callbacks for manual recognition. Then, their keys can be used to call the service.

How does manual recognition determine whether content is non-compliant?

After content is pushed to manual recognition, the message called back contains the `ManualDetailCode` field, which represents the type of violation identified by manual recognition.

Can the QPS of manual recognition be adjusted?

Yes. You can submit a ticket to apply for the adjustment.

I don’t see the data I want in the AMS console.

Please check the following:

Make sure the currently logged-in account is the account that called the API.

The console data is displayed at the account level, which means that you can only see the data under the currently logged in account, no matter it’s a root account or a sub-account.

The console displays the T+1 day data; that is, you can only view the data of a day in the console on the next day.

Why can't a sub-account see the data under the root account and other sub-accounts?

The CMS data in the console under the root account and sub-accounts as well as data under different sub-accounts are independent of each other. If a sub-account wants to view data under the root account or other sub-accounts, it needs to be authorized by the root account by associating it with the `QcloudCamReadOnlyAccess` policy on the Policies page in the CAM console.

Can I configure a threshold for a risk model in the console?

No. You can submit a ticket for assistance with configuration on the backend.

Can I modify the default QPS (queries per second)?

You can submit a ticket for assistance with modification on the backend.

Is the QPS counted on the account level or Biztype level?

QPS is counted on the account level and defaults to 20 QPS.

Can I specify Biztype on my own?

Yes, but you need to submit a ticket for it first. `Biztype` can contain 3 - 32 letters, digits, and underscores. It can not be changed once configured.

Will the `Review` field output for each tag?

No. It will be output for only tags with a score output by auto review but not for such tags as ad and QR code.

Can I add types of malicious content by myself?

No.

The custom dictionary doesn't take effect. What should I do?

You can troubleshoot the problem as follows:

After creating the custom dictionary, you need to create a custom policy and bind the dictionary to the policy for the dictionary to take effect. After configuring the policy, enter the policy name as the Biztype input parameter during API call, and the policy will take effect.

If you use the default policy for testing purposes, you need to add custom keywords to the non-compliant word dictionary on the “Custom Library Management > Preset Dictionary” page for them to take effect.

If the problem persists, submit a ticket for assistance.

I have already activated the AMS service, but when I call the short audio API, the system prompts that I don't have access to the API. What should I do?

The short audio API has not been opened up, so it cannot be called.

When I call the audio moderation task creation API of AMS, the system prompts that "can not access bucket tianyu-content-moderation-1305692660 at region ap-guangzhou – write to bucket error The specified bucket does not exist." What should I do?

After activating the service in the CMS console, you will be prompted to authorize a COS bucket. Specifically, a bucket will be created to store the segments generated during moderation.

The above error will occur if you delete the COS bucket by mistake. You can submit a ticket and provide the `RequestId` for assistance.

If I set two or more keys, which one should I use?

You can use any Tencent Cloud key, as the keys are shared; that is, they can be used for one or more services.