Sightengine

What is Sightengine?

Sightengine is a content moderation API provider that analyses images, videos, text and audio to detect unwanted, unsafe or synthetic content. It is used by platforms that accept user-generated content and need an automated first line of defence before anything is published.

Its catalogue covers more than 100 moderation classes, including nudity, violence, weapons, drugs, self-harm, hate symbols and scams. It also provides a dedicated family of models for synthetic media, including deepfake detection and AI-generated content detection.

This second family of models makes Sightengine relevant beyond classic moderation. It answers “Is this real?” rather than only “Is this allowed?” On Eden AI, Sightengine is available for image deepfake detection. The rest of its moderation stack is accessed directly through Sightengine.

Sightengine at a Glance

Sightengine’s Main AI Capabilities

Sightengine’s own product range covers:

• Deepfake detection: Face swaps, face regeneration and AI-modified faces inside otherwise real media
• AI-generated image and video detection: Content produced entirely by a generative model, with per-generator scores
• Nudity and sexual content detection: Graded classes rather than a single binary flag
• Violence, weapons, drugs and self-harm detection
• Text moderation: Profanity, hate speech, personal data, scams and grooming patterns
• Audio and video moderation: Frame-level and stream analysis
• Image quality and property analysis: Blur, quality, colours, faces, celebrities and logos

If you need the full moderation suite in one place, contact the Eden AI team. Provider coverage is expanded based on customer demand.

When Should You Choose Sightengine?

Sightengine is the right choice when your problem is authenticity, not just compliance. It helps you determine whether the face in an uploaded image has been manipulated or synthetically produced.

Choose Sightengine on Eden AI when:

• Users upload profile pictures, selfies or identity photos and you need to detect face swaps or AI-generated faces
• You publish or resell visual content and need a verification step before it goes live
• You run a fraud, KYC or account-integrity pipeline and want a synthetic-media signal alongside your existing checks
• You need EU-hosted processing for GDPR-related reasons
• You prefer per-image pricing instead of a monthly moderation subscription

Sightengine is not the right choice when:

• You need OCR, speech-to-text, translation or text generation - use the dedicated providers available on Eden AI
• You specifically need audio or video deepfake detection - the model currently exposed on Eden AI is image-based
• You need a human-readable explanation of why something was flagged - Sightengine returns a score, not a forensic report

Sightengine Pros and Cons

Pros

• Purpose-built for synthetic media: Deepfake detection is a first-class model rather than a secondary feature of a general-purpose vision API
• Simple, interpretable output: You receive a score between 0 and 1 and choose the threshold yourself
• Predictable pricing: $0.015 per image on Eden AI, with no Sightengine subscription
• EU hosting: Important when submitted images contain personal data
• Strong adjacent catalogue: Useful if you later expand into nudity, violence or text moderation

Cons

• Narrow scope on Eden AI: Only one image model is currently available
• Threshold management: The score-based output requires you to define and maintain your own thresholds
• Variable detection quality: Accuracy can vary depending on image resolution, compression and how recently the image generator was released
• No like-for-like fallback: There is currently no equivalent provider for the same task on Eden AI
• Different direct pricing model: Direct Sightengine plans start at approximately $29 per month, so comparing billing options may be useful at high volumes

Sightengine Models, Features and Capabilities on Eden AI

Available Sightengine Models on Eden AI

Deepfake Detection vs. AI-Generated Content Detection

These two technologies are often confused. Choosing the wrong model can lead to unnecessary or ineffective API calls.

When your primary risk is impersonation, deepfake detection is the appropriate model. When your primary risk is undisclosed synthetic content, use AI-generated content detection. For stricter verification pipelines, both models can be used together.

Sightengine API Output: What You Receive

The higher the score, the more confident the model is that the image contains a manipulated or synthetic face. There is no predefined verdict field. Your team must define the decision boundary.

Choosing a Threshold

A single global threshold is rarely appropriate. A three-band approach is usually more effective in production:

• Low score: Publish the image automatically
• Middle band: Send the image for human review
• High score: Block the image or require an additional verification step

Begin by scoring several hundred images for which you already know the correct answer. Review where your false positives and false negatives occur, then define two thresholds. Reassess these thresholds every few months because image-generation systems evolve faster than fixed moderation rules.

Important Note on Sightengine Accuracy and Reliability

Detection accuracy is not a fixed number. It varies depending on:

• Image resolution
• Compression
• Whether the image has been re-uploaded
• Whether the image is a screenshot
• Face size and angle
• How recently the manipulation technology was released

Benchmark Sightengine using your own traffic before connecting its score to an automated blocking decision. Use screenshots and compressed images from your actual application rather than relying only on clean research samples.

Supported AI Categories

• Vision
• Image Analysis
• Content moderation

What Can You Build with Sightengine?

Use Case 1: Screen Profile Pictures and Avatars at Signup

Score every avatar when it is uploaded. Automatically approve images with low-risk scores, hold suspicious images for review and prevent face-swapped or AI-generated identities from reaching your community. This is one of the most valuable placements for the model because verification happens before the content becomes public.

Use Case 2: Verify User-Generated Images Before Publication

Marketplaces, dating applications, forums and review platforms can add a deepfake check to the same pipeline that already handles nudity and violence detection. With one additional call per image, manipulated media becomes part of the automated moderation process rather than a manual discovery problem.

Use Case 3: Add a Synthetic-Media Signal to Fraud and KYC Flows

Identity documents and selfie checks are increasingly targeted by deepfake technology. Scoring a submitted image gives your fraud engine an additional signal that can be evaluated alongside device, behavioural and document-verification signals. This signal is valuable because it is independent of the other checks.

Use Case 4: Support Editorial and Fact-Checking Review

Newsrooms and brand-safety teams can use risk scores to prioritise incoming visual content. Instead of manually examining every image, reviewers can focus their attention on images with the highest probability of manipulation.

Sightengine Use Cases by Industry

Why Use Sightengine Through Eden AI?

One API for Sightengine and More Than 50 AI Providers

Sightengine is available through the same interface as OCR, speech, language model and computer vision providers on Eden AI. A moderation pipeline requiring a deepfake score, an explicit-content check and an OCR analysis can make three calls using one API and receive one consolidated bill. This avoids managing three separate vendor relationships.

Combine Sightengine with Other Moderation Models

Deepfake detection answers one specific question. Combine it with explicit-content detection, face detection or AI-generated image detection from other Eden AI providers to cover both authenticity and safety in a single workflow. Each model can then focus on the task for which it is designed.

Monitor Usage, Billing and Costs in One Place

Use the Eden AI dashboard to track:

• API calls
• Success rates
• Latency
• Spending by provider

Because deepfake detection is priced per image, your costs will generally follow your image-upload volume. This is especially important during marketing campaigns and traffic spikes.

Keep Your Options Open

Sightengine is currently the deepfake detection provider available on Eden AI. Because your application connects to the Eden AI endpoint rather than directly to Sightengine, adding or switching to another provider in the future can be handled through a parameter change rather than a full reintegration.

How to Integrate Sightengine with Eden AI

Integration Overview

1. Create an Eden AI account and obtain your API key
2. Select the Image Deepfake Detection feature
3. Set sightengine as the provider
4. Send an image URL or upload an image file
5. Read the deepfake score and apply your thresholds
6. Monitor costs, latency and error rates in the Eden AI dashboard

Example Request

import requests

url = "https://api.edenai.run/v3/universal-ai/"

headers = {
    "Authorization": "Bearer <your-api-key>",
    "Content-Type": "application/json",
}

payload = {
    "model": "image/deepfake_detection/sightengine",
    "input": {
        "file": "<file_id_from_upload_endpoint>",
    },
    "show_original_response": False,
}

response = requests.post(
    url,
    headers=headers,
    json=payload,
    timeout=30,
)

response.raise_for_status()

data = response.json()
print(data)

‍

Authentication

Keep your Eden AI API key on your server. Never call the endpoint directly from a browser or mobile application with the API key embedded. Proxy the request through your own backend and rotate the key immediately if it is exposed.

Response Format

Validate the response before taking action:

• Check the provider status field
• Confirm that the score is present
• Explicitly handle unsuccessful requests

If the score is missing, send the image to human review rather than automatically allowing it.

Production Best Practices

• Test the model using your own images, including real uploads and realistic compression
• Use a human-review band rather than one hard threshold
• Configure timeouts and retries with exponential backoff
• Do not retry a paid request indefinitely
• Cache results using an image hash
• Log the model score alongside the final human decision
• Repeat benchmarks every few months as new image generators appear

Best Sightengine Alternatives and Comparisons on Eden AI

Sightengine vs. Other Image Moderation Providers on Eden AI

Sightengine is currently the only provider exposing deepfake detection on Eden AI. As a result, a direct comparison with an equivalent provider inside the platform is not currently possible. However, you can compare providers for the wider content-moderation workflow.

Most production moderation systems use Sightengine for authenticity checks and a general-purpose vision provider for safety categories rather than choosing one instead of the other.

Sightengine vs. Hive AI and Reality Defender

Outside Eden AI, the closest competitors for synthetic-media detection include:

• Hive AI
• Reality Defender
• Sensity
• Resemble AI

These providers primarily differ in:

• Supported media formats, such as images, video and audio
• The level of detail in their reports
• Their commercial models

Many enterprise-focused providers use quote-based pricing rather than publishing standard rates. Sightengine through Eden AI offers a published per-image price that can be used immediately. For a complete comparison, see Eden AI’s deepfake detection API comparison.

Frequently Asked Questions About Sightengine on Eden AI