AI Guardrails:Voice Moderation 2.0 multilingual service

How it works

Submit an audio or video URL to start an asynchronous moderation task. The service splits the audio into fixed-length segments, runs multiple parallel detection models against each segment, and returns:

• The transcribed text for every segment

• One or more risk tags (e.g., profanity, contraband) per segment

• A per-segment risk level (high, medium, low, or none)

• A temporary URL to the audio clip for manual review

Retrieve results by polling the query API or by configuring a callback URL to receive push notifications when the task completes.

What's new in 2.0

In Voice Moderation, queries per second (QPS) refers to the number of API requests processed per second. Concurrent streams refers to the number of audio files or live streams being analyzed simultaneously.

Supported regions and languages

For 35-language detection, use the Singapore region. The US (Virginia) region supports Chinese and English only.

Supported formats

Audio files: MP3, WAV, AAC, WMA, OGG, M4A, AMR (up to 500 MB)

Video files: AVI, FLV, MP4, MPG, ASF, WMV, MOV, RMVB, RM (up to 500 MB)

Live stream protocols: RTMP, HLS, HTTP-FLV, RTSP

Risk tags

When content contains multiple risk types, the service returns all matching tags simultaneously. Tags are returned in the labels field; sub-tags in the riskTips field.

Billing

Voice Moderation 2.0 uses pay-as-you-go billing. Charges are calculated daily based on actual usage. There is no charge for days when the service is not used.

Fees apply only to requests that return HTTP 200. Requests that return other status codes are not charged.

Prerequisites

Before you begin, make sure you have:

• An Alibaba Cloud account with Voice Moderation 2.0 activated

• A RAM user with the AliyunYundunGreenWebFullAccess policy attached

• An AccessKey pair for the RAM user

Activate the service

Go to Activate Service to activate Voice Moderation 2.0.

Set up RAM permissions

1. Log in to the RAM console as a RAM administrator.

2. Create a RAM user. For details, see Create a RAM user.

3. Attach the AliyunYundunGreenWebFullAccess system policy to the RAM user. For details, see Grant permissions to a RAM user.

The RAM user can now call Content Moderation API operations.

Install the SDK

Use the software development kit (SDK) to call the API without constructing raw HTTP requests. For SDK sample code in languages not listed below, use the OpenAPI Developer Portal to auto-generate code for the VoiceModeration operation.

For SDK installation and integration details, see Voice Moderation 2.0 SDKs and integration guide.

API reference

The service endpoint pattern is https://green-cip.{region}.aliyuncs.com.

Two API operations are available:

• VoiceModeration — Submit a moderation task

• VoiceModerationResult — Query a moderation task result

For information about constructing raw HTTP requests, see Make a raw HTTP call.

{
  "service":"audio_multilingual_global",
  "serviceParameters":"{\"taskId\":\"AAAAA-BBBBB"}"
}

{
  "service":"audio_multilingual_global",
  "serviceParameters":"{\"cryptType\":\"SHA256\",\"seed\":\"abc***123\",\"callback\":\"https://aliyun.com/callback\",\"url\":\"http://aliyundoc.com/test.flv"}"
}

{
  "code": 200,
  "data": {
    "taskId": "AAAAA-BBBBB"
  },
  "message": "SUCCESS",
  "requestId": "AAAAAA-BBBB-CCCCC-DDDD-EEEEEEEE****"
}

{
  "Code": 200,
  "Data": {
    "DataId": "voice20240307***",
    "LiveId": "liveId1****",
    "RiskLevel": "high",
    "SliceDetails": [
      {
        "StartTime": 0,
        "EndTime": 4065,
        "Labels": "political_content,xxxx",
        "RiskLevel": "high",
        "RiskTips": "contraband_ProhibitedGoods",
        "RiskWords": "Risk Word A",
        "Text": "AI Guardrails product test case",
        "Url": "https://aliyundoc.com"
      }
    ]
  },
  "Message": "OK",
  "RequestId": "AAAAAA-BBBB-CCCCC-DDDD-EEEEEEEE****"
}