Provider

Resemble AI

Detects deepfakes and AI-generated content across images and video.

summary

Detects deepfakes and manipulated media in images and video with DETECT-World, Resemble AI's latest detection model.
Supports fraud prevention, impersonation detection, content moderation, and media verification workflows.
Available through Eden AI's unified API alongside other detection providers, so you can compare results or switch without rewriting your integration.

Resemble AI is a generative AI security company founded in 2019 and based in Mountain View, California. It helps organizations detect deepfakes, AI-generated content, and manipulated media. Through Eden AI, developers can access Resemble AI's multimodal detection technology, including its latest model DETECT-World, through one unified API - with the option to route the same request to another detection provider for comparison or fallback.

Provider name	Resemble AI
Main category	Deepfake Detection, AI Content Detection
Technologies	Deepfake Detection, AI Content Detection
Main detection technology	DETECT-World (DETECT-3B Omni also available)
Available through Eden AI	Deepfake Detection (image), Deepfake Detection (video, asynchronous), AI Content Detection (image)

Main AI Capabilities

Deepfake Detection: Analyze images and video for signs that content has been artificially generated or manipulated.
AI Image Detection: Identify signals associated with AI-generated imagery.

Resemble AI's native platform also includes audio deepfake detection, voice cloning, speech-to-speech conversion, and watermarking. These are not currently exposed through the Eden AI integration. Consult the Eden AI documentation or dashboard for the current list of supported features.

When Should You Choose Resemble AI?

Fraud prevention: Analyze suspicious images or videos submitted during account creation, payments, claims, or identity-verification processes.
Impersonation detection: Detect media that may imitate executives, employees, customers, creators, or public figures.
Trust and safety: Add synthetic-media checks to content platforms, social networks, marketplaces, and user-generated-content workflows.
Media verification: Analyze images and videos before publishing or distributing them.
Content moderation: Screen uploaded media for indicators of AI generation before escalating suspicious content for human review.
Coverage of new generators: Choose Resemble AI when your threat model includes generators released faster than detection datasets can be built for them.

Resemble AI pros and cons

Pros	Cons
Detects AI-generated and manipulated images and video	Detection results should not be treated as absolute proof
One model covers several media types	Performance can vary depending on compression and media quality
Benchmarked against more than 250 generative models	High-risk decisions may still require human review
Physical-consistency checks extend coverage to unseen generators	Image and video benchmarks are self-reported, pending third-party validation
Same API across detection models, no migration required	Detection thresholds must be tested for each use case
Relevant for fraud, moderation, and trust and safety	Audio detection is not currently available through Eden AI

Resemble AI Models, Features, and Capabilities

Choosing the right detection setup means matching your media types, latency tolerance, and risk level to the model that fits. The right choice depends on how much of your traffic is image versus video, how quickly you need a verdict, and what a false positive costs you.

What Is DETECT-World?

DETECT-World is Resemble AI's most recent detection model, released in August 2026. It builds on DETECT-3B Omni rather than replacing it: every input is evaluated on two dimensions at once. The first is artifact analysis, which looks for the statistical traces that generation models leave in raw pixel and audio data. The second is a physical-consistency layer that assesses whether elements such as lighting and motion behave the way they would in a real scene.

The purpose of the second layer is coverage of generators the model has not been trained on. Detection systems that rely only on recognising known generators can see false negatives rise in the window before they are updated for a newly released tool. Because physical consistency is not tied to any specific generator's fingerprint, DETECT-World can flag unfamiliar sources while retraining catches up.

Resemble AI reports benchmarking against more than 250 generative models and support for 54 languages, validated against MLAADv10.

Reported accuracy: image 95.77% and video 98.2%, both internal benchmarks pending third-party validation. Resemble AI also reports 99.47% on audio, externally benchmarked via Podonos, though audio detection is not currently exposed through Eden AI. As with any detection model, test against your own content and traffic before setting production thresholds.

What Is DETECT-3B Omni?

DETECT-3B Omni is Resemble AI's multimodal model for detecting AI-generated and manipulated content across audio, images, and video. It is the foundation DETECT-World is built on, and remains available.

The model has been tested against more than 160 generative models and supports more than 50 languages and a wide range of media formats. Its multimodal design allows teams to analyze different content types through one detection system instead of maintaining separate tools for image and video verification.

Available Resemble AI Models

The models available through Eden AI may depend on the selected feature. Consult the Eden AI model catalog to verify which models are currently accessible.

Model	Purpose	Available on
DETECT-World	Latest multimodal detection model, combining artifact analysis with physical-consistency checks. Served through the same Resemble Detect API, no integration change required	Image and video detection endpoints
DETECT-3B Omni	Multimodal deepfake and AI-content detection. The foundation DETECT-World is built on	Image and video detection endpoints

Resemble AI API Output: what data can be extracted or generated?

Feature	Eden AI path	Input	Output
Deepfake Detection — Image	image/deepfake_detection/resemble	Image file	Detection verdict and confidence score
AI Content Detection — Image	image/ai_detection/resemble	Image file	Signals indicating whether the image may be AI-generated
Deepfake Detection — Video	video/deepfake_detection_async/resemble	Video file	Detection verdict and confidence score, returned asynchronously

What Can You Build With Resemble AI?

Fraud-detection systems

Analyze suspicious media submitted during account registration, insurance claims, financial transactions, or customer-support interactions.

Identity-verification workflows

Add deepfake checks to image and video identity-verification processes, including document photos and liveness captures.

Impersonation detection tools

Detect content that may imitate employees, executives, creators, or public figures.

Content-moderation platforms

Screen user-uploaded images and videos for signs of synthetic generation or manipulation.

Media-verification tools

Help journalists, platforms, and organizations assess whether submitted media may have been artificially generated.

Multi-provider verification

Route the same media to Resemble AI and another detection provider through one Eden AI request, and treat disagreement between them as a signal to escalate for human review.

Resemble AI Use Cases by Industry

Industry	Use Cases
Financial Services	Fraud prevention, deepfake detection in KYC, and identity verification
Insurance	Fraudulent-claim detection and submitted-media verification
Telecommunications	Impersonation detection and account-takeover prevention
Marketplaces	Profile-image verification and synthetic-content screening
Social Media	Content moderation and deepfake detection
Media and Publishing	Source verification and manipulated-media analysis
Public Sector	Misinformation investigations and media authentication
Entertainment	Content protection and unauthorised-likeness detection

How to Integrate Resemble AI With Eden AI

Create an Eden AI account.
Generate an Eden AI API key.
Choose the required feature:
- image/deepfake_detection/resemble
- image/ai_detection/resemble
- video/deepfake_detection_async/resemble
Select Resemble AI as the provider.
Submit the relevant audio, image, video, or text input.
Retrieve the result using the response format documented for the selected feature.
Monitor API usage, latency, cost, and detection outcomes from your Eden AI workspace.

Frequently Asked Questions about Resemble AI on Eden AI

What is Resemble AI used for?

Resemble AI is used to detect deepfakes, AI-generated content, and manipulated media. Through Eden AI, its image and video detection capabilities are available via a unified API.

Can Resemble AI detect image and video deepfakes?

Yes. DETECT-World and DETECT-3B Omni are multimodal detection models designed to analyze images and video for signals associated with AI generation or manipulation.

What is DETECT-World?

DETECT-World is Resemble AI's latest deepfake detection model, released in August 2026. It combines the artifact-based analysis used in DETECT-3B Omni with a physical-consistency layer that checks whether lighting and motion in a scene behave realistically, allowing it to flag content from generators it has not been trained on. It has been benchmarked against more than 250 generative models.

What is the difference between DETECT-World and DETECT-3B Omni?

DETECT-World is built on DETECT-3B Omni and adds a second detection method on top of it. Both cover images and video. DETECT-World runs artifact analysis and physical-consistency checks together, so content that evades one check can still be caught by the other. DETECT-3B Omni remains available through Eden AI.

Do I need to change my integration to use DETECT-World?

No. DETECT-World is served through the same Resemble Detect API, so no endpoint or code change is required. If you have configured confidence thresholds based on earlier results, test them again with representative content, as score distributions can differ between detection models.

Which Resemble AI endpoints are available on Eden AI?

Three: image/deepfake_detection/resemble for image deepfake detection, image/ai_detection/resemble for AI-generated image detection, and video/deepfake_detection_async/resemble for asynchronous video deepfake detection.

Is audio deepfake detection available?

Audio detection is part of Resemble AI's native platform but is not currently exposed through Eden AI. Check the Eden AI model catalog for the latest availability.

How accurate is Resemble AI deepfake detection?

Resemble AI reports 95.77% on images and 98.2% on video from internal benchmarks pending third-party validation. Real-world performance depends on media format, quality, compression, and manipulation method. Results should be treated as one verification signal and combined with human review or additional checks in high-risk workflows.

Do I need a Resemble AI API key?

No. When using Resemble AI through Eden AI, you authenticate with your Eden AI API key. Eden AI manages access to the underlying provider integration.