Skip to main content

​
Introduction

Detectors are the building blocks of judges. Each detector analyzes specific aspects of content and can be configured with custom parameters.

​
Configuration Format

When creating a judge using the v1 endpoint, detectors are configured in the detectors array: 
{
  "name": "My Judge",
  "description": "Judge description",
  "content_type": "video",
  "detectors": [
    {
      "detector_name": "video_specs",
      "config": {
        "min_width": 1920,
        "min_height": 1080,
        "min_fps": 24
      }
    },
    {
      "detector_name": "audio_noise",
      "config": {}
    }
  ]
}

​
Available Detectors

​
Content Matching Detectors

DetectorDescription
audio_matchingChecks whether audio matches reference audio
splitscreen_video_matchingChecks split-screen content against references or descriptions
video_matchingChecks whether video matches reference videos
video_matching_descriptionChecks whether video content matches a description

​
Content Analysis Detectors

DetectorDescription
languageChecks detected language against an allowed list
sentimentChecks sentiment and prohibited content in transcripts
title_and_descriptionChecks title and description for required elements
transcript_policyChecks transcripts for required and prohibited language
transcript_productCounts brand and product mentions in transcripts
clipper_demographicsExtracts demographic data from social media screen recordings

​
Video Visual Analysis Detectors

DetectorDescription
backgroundAssesses background conditions in the video
framingChecks whether framing meets composition expectations
lightingEvaluates lighting balance across video frames
overlaysChecks for required or disallowed overlays
product_applicationChecks whether a product is being applied
subtitlesChecks for subtitle presence
transition_detectionCounts scene transitions
video_qualityEvaluates overall video quality
video_specsValidates video dimensions and duration
wardrobeChecks wardrobe and beauty rule compliance

​
Image Visual Analysis Detectors

DetectorDescription
image_ip_infringementFlags potential IP concerns in images
image_sharpnessEvaluates image sharpness
image_specsValidates image dimensions and aspect ratio

​
Audio Analysis Detectors

DetectorDescription
audio_noiseEvaluates background noise and speech clarity
pronunciationEvaluates brand and product pronunciation
voice_overDetects voice-over narration

​
Reference Content

Some detectors require reference content (for example, reference videos or audio) for comparison. Reference content must be uploaded to Google Cloud Storage and provided as gs:// URIs.

​
Detectors Requiring Reference Content

  • audio_matching - Requires reference_audio
  • video_matching - Requires reference_videos
  • splitscreen_video_matching - Requires at least one of top_content_references or bottom_content_references
  • pronunciation - Optional brand_audio_url and product_audio_url

​
Uploading Reference Content

  1. Use the /v1/judges/reference_content endpoint to get a pre-signed upload URL
  2. Upload your reference file to the provided URL
  3. Use the returned reference_content_uri in your detector configuration

​
Configuration Validation

  • Detector names must be valid (from the list above)
  • Required configuration keys must be provided for each detector
  • Reference content URIs must use the gs:// scheme
  • File extensions for reference content must match allowed types

​
Next Steps

  • See individual detector documentation for detailed configuration options
  • Check the Create Judge endpoint for full examples
  • Review detector-specific configuration schemas for required and optional parameters