AI-Generated and Deepfake Content Detection Playground

Hive's AI-Generated and Deepfake Content Detection API

Hive’s AI-Generated & Deepfake Content Detection API uses deep learning to detect AI-generated visuals, visual deepfakes, and (when present) AI-generated audio.

It can also attribute likely generator models (e.g., Midjourney, Stable Diffusion).

Each category is represented by a model head. Heads contain mutually exclusive classes whose confidence scores sum to 1 (per head).

Our Playground supports image AND video inputs. For videos, Hive splits the input into frames and returns timestamped predictions per frame.

How It Works

When an image or video is submitted, Hive returns an array of classification results. Each result includes a list of classes, each with a class and a model-generated confidence value between 0 and 1.

Example structure:

{
  "output": [
    {
      "extra": [
        { "name": "frame_index", "value": 0 },
        { "name": "timestamp",   "value": 0.0 }  // Always 0.0 for images
      ],
      "classes": [
        { "class": "not_ai_generated", "value": 0.98 },
        { "class": "ai_generated",     "value": 0.02 },

        // AI Generator attribution head (subset shown)
        { "class": "midjourney",       "value": 0.01 },
        { "class": "stablediffusion",  "value": 0.00 },
        { "class": "none",             "value": 0.99 },

        // Deepfake (visual) head
        { "class": "deepfake",         "value": 0.03 },

        // Audio head (if input has audio)
        { "class": "not_ai_generated_audio", "value": 1.0 },
        { "class": "ai_generated_audio",     "value": 0.0 }
	// ai_generated_audio is always 0.0 for images and  videos with no audio
      ]
    }
  ]
}

Note: inconclusive and inconclusive_video are also returned in the classes array, but these are legacy fields and will always have value 0. They are unused.

Quickstart

curl --request POST \
  --url 'https://api.thehive.ai/api/v3/hive/ai-generated-and-deepfake-content-detection' \
  --header 'authorization: Bearer <YOUR_SECRET_KEY>' \
  --header 'Content-Type: application/json' \
  --data '{
    "media_metadata": true,
    "input": [
      { "media_url": "https://i.ibb.co/DHzCRsNk/Screenshot-2025-06-05-at-3-25-40-PM.png" }
    ]
  }'

import requests, json

API_KEY = "<YOUR_SECRET_KEY>"
url = "https://api.thehive.ai/api/v3/hive/ai-generated-and-deepfake-content-detection"

payload = {
  "media_metadata": True,
  "input": [
    { "media_url": "https://i.ibb.co/DHzCRsNk/Screenshot-2025-06-05-at-3-25-40-PM.png" }
    # OR: { "media_base64": "data:<mime>;base64,<BASE64_DATA>" }
  ]
}

res = requests.post(
  url,
  headers={"Authorization": f"Bearer {API_KEY}", "Content-Type": "application/json"},
  data=json.dumps(payload),
  timeout=60
)
print(res.status_code, res.json())

const API_KEY = "<YOUR_SECRET_KEY>";

const res = await fetch(
  "https://api.thehive.ai/api/v3/hive/ai-generated-and-deepfake-content-detection",
  {
    method: "POST",
    headers: {
      "authorization": `Bearer ${API_KEY}`,
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      media_metadata: true,
      input: [
        { media_url: "https://i.ibb.co/DHzCRsNk/Screenshot-2025-06-05-at-3-25-40-PM.png" }
        // OR: { media_base64: "data:<mime>;base64,<BASE64_DATA>" }
      ],
    }),
  }
);

console.log(await res.json());

Tip: For large files, you can also upload using multipart/form-data with --form 'media=@"/abs/path/file.ext"'.

Generate your V3 API Key:

1. Click ‘Service API Keys’ in the sidebar.
2. Click ‘Create API Key’ to create a new key scoped to your organization. This key can be used with any "Playground Available" model.

⚠️ Important: Keep your API Key secure. Do not share it publicly.

V3 (Playground) vs. V2 (Enterprise)

• V3 (Playground)
  • Easiest way to try the model and integrate quickly.
  • Default rate limit: 100 requests/day.
  • Works with the endpoint used in the examples above.
• V2 (Enterprise Projects)
  • For higher rate limits and dedicated stats.
  • You’ll receive a project-specific API key and use our V2 task endpoints (sync/async).
  • Contact us to enable a V2 Enterprise project!

Once your V2 project is enabled, we’ll provide the exact endpoints and keys inside Projects. Your threshold logic and class names remain the same across V3/V2.

Request Format

To send an image or video to our AI-Generated and Deepfake Content Detection API, make a POST request to the endpoint below. You can pass in a public image, video, or audio URL using the media_url parameter.

Supported Media Types

Image: .jpg, .png, .webp, .gif

Video: .mp4, .webm, .m4v

Audio: .mp3, .m4a, .wav

Using media_url or media_base64

curl --location 'https://api.thehive.ai/api/v3/hive/ai-generated-and-deepfake-content-detection' \
--header 'authorization: Bearer <YOUR_SECRET_KEY_HERE>' \
--header 'Content-Type: application/json' \
--data '{
  "input": [
    {
      "media_url": "https://hivemoderation.com/images/31f7d53.png"
      // OR you could use "media_base64": "<your base64 encoding here>"
    }
  ]
}'

Using Multipart form

For images and videos that are too large for a base64 JSON body, you can send them as a standard multipart/form-data Content Type upload.

curl --location 'https://api.thehive.ai/api/v3/hive/ai-generated-and-deepfake-content-detection' \
--header 'authorization: Bearer <YOUR_SECRET_KEY_HERE>' \
--header 'Content-Type: multipart/form-data' \
--form 'media=@"<absolute path to your file>"'

Response Format

When you submit an image or video, the API returns a JSON object with task metadata and one or more frame-level results.

Class categories you'll see:

• AI-generated authenticity (image/per frame): ai_generated, not_ai_generated
• Deepfake (image/per frame): deepfake
• AI Generator attribution (image/per frame): Many engine labels (e.g. midjourney, stablediffusion, dalle, etc.) plus other_image_generators and none (no specific generator detected)
• Audio authenticity (videos w/ audio, or audio-only inputs): ai_generated_audio, not_ai_generated_audio
  • Images and videos w/o audio will default to ai_generated_audio == 0.

Thresholds

We recommend the following thresholds for optimized model performance:

1. AI-Generated Images: ai_generated ≥ 0.9
2. AI-Generated Videos: ai_generated ≥ 0.9 on any frame
3. Deepfake Image: deepfake ≥ 0.9

Need More Help?

Contact us if you need help getting set up with AI-Generated and Deepfake Content Detection!