Scope before scores

Synthetic Media Detection

Understand the difference between synthetic scenes, manipulated faces, and edited photos. FauxGuard checks still-image signals only; it does not analyze video, audio, or live streams.

Still images only. One-time purchase, no subscription or daily free allowance. Save your private recovery code before paying.

$4.99 for 20 image checks

One-time purchase. No subscription. No account required.

Loading your balance…

Already purchased? Restore your checks

Drop a photo here

JPG, PNG, or WebP. Each completed model analysis uses one image check.

Synthetic media is a category, not one detection task

Synthetic media can include generated images, manipulated likenesses, artificial speech, and altered video. These outputs may share a generative workflow, but they require different evidence to investigate. A still-image model does not inspect voice characteristics or changes across video frames. Before choosing a tool, define which medium you have and which claim needs verification. Otherwise, a useful answer about one file may be misapplied to a much larger question.

FauxGuard's scope on this page is deliberately narrow: JPG, PNG, and WebP still images. The upload area uses the existing image detector, not a separate video or audio service. It requests AI-generation and face-manipulation scores and reads available clues in the processed file. It cannot certify the authenticity of a recording, detect a cloned voice, or verify that a particular person made a statement depicted in a screenshot.

Separate a generated scene from an edited photograph

A scene created by a model and a camera photograph altered with generative fill may both involve AI, but not in the same way. In the second case, most of the image may still come from a camera. A global score cannot reconstruct every operation or locate a small changed region. The useful review asks what part of a claim depends on the image and what evidence would distinguish the plausible explanations.

Ordinary editing adds another distinction. Cropping, exposure correction, skin retouching, and compositing are not automatically equivalent to AI generation. An editor name in metadata can suggest a workflow without identifying the exact action. A low AI-generation score does not rule out manual edits, while a high score does not establish deceptive intent. Synthetic media can be clearly labeled creative work, just as a real photograph can be attached to an inaccurate caption.

What the still-image detector actually returns

FauxGuard reports two visual model scores separately. The AI-generation score addresses model patterns associated with generated content. The face-manipulation score addresses a different signal and is not an identity match. Confidence labels group score ranges; they are not calibrated probabilities or an independently validated accuracy rate. A report should preserve those distinctions instead of compressing them into a single claim that an image is genuine or fake.

Readable file findings can include camera fields, a software tag, or a generator name. Those details are unverified and may be copied, changed, or missing. Browser conversion can remove original metadata before the server reads the processed image. FauxGuard does not authenticate C2PA signatures, verify SynthID watermarks, return a local tamper heat map, or establish an edit timeline. When the model cannot supply complete scores, a partial notice appears and no image check is used.

Why a frame cannot authenticate an entire video

You can upload a supported still frame from a video, but the result describes that image alone. It does not evaluate movement across frames, synchronization with sound, or the rest of the clip. A frame may look ordinary while the recording contains a misleading audio track, and an unusual frame may be caused by blur or compression. Do not describe one frame's score as a verified conclusion about the complete recording.

Preserve the original clip and its publication context if a video claim matters. Record which frame you checked, including an approximate position in the clip, without assuming it represents every moment. Seek a workflow designed for the relevant medium when the unresolved question concerns audio or temporal changes. FauxGuard does not perform that additional analysis, and purchasing image checks does not add support for video, live streams, or audio uploads.

  • Still image: evaluate model signals and file clues in the submitted image.
  • Video: preserve the full recording; a frame check covers only the frame.
  • Audio: use an appropriate audio review process; this upload does not accept recordings.
  • Source claim: investigate publication context independently of model scores.

Use synthetic-media evidence without overstating it

A content reviewer may need to decide whether to ask for an original file, look for earlier versions, or escalate a concern. A detector can support that triage, but it should not be the sole basis for accusing someone, denying a claim, or asserting that a public statement is false. Document the observation, the model inference, and the unanswered questions separately. The consequences of a decision should determine how much independent evidence is required.

For a still-image question, start with the upload area above and interpret the result alongside source research. For a miscaptioned news photo, an earlier publication may matter more than a model score. For a suspicious portrait, the face-manipulation guide explains why an identity claim needs separate verification. An unresolved answer is preferable to claiming certainty beyond what the tool measured. The guides below provide practical next steps without implying broader media support than is actually available.

Detector

Check an image with FauxGuard

Upload a JPG, PNG, or WebP image and review separate model scores, score-based confidence, readable file clues, and reasons. No account is required. A score is evidence to review, not proof of origin.

Related image checks

Frequently asked questions

Does this synthetic media detector accept video or audio?

No. FauxGuard currently accepts JPG, PNG, and WebP still images only. You may check an extracted frame as an image, but its result does not authenticate the clip, sound, or live stream. Image checks do not unlock video or audio analysis.

Is all synthetic media deceptive?

No. Generated illustrations, fictional scenes, and clearly disclosed creative edits can be legitimate uses. A model signal concerns patterns in the submitted file; it does not establish intent, consent, publication context, or the truthfulness of a caption. Investigate those questions separately.

Do I need an account to check an image?

No account or login is required. Your image-check balance is accessed through a private recovery code and a browser cookie. Save the code before paying and keep it private: anyone with it can spend your checks. You can restore the balance from the detector or pricing page on another device.

How much does an image check pack cost?

FauxGuard sells 20 image checks for $4.99 USD as a one-time purchase. There is no subscription or daily free allowance. One completed model analysis uses one check. Failed model requests and metadata-only fallback results do not use a check. Review pricing and the terms before paying.

How should I read the detection result?

Read the AI-generation score and the separate face-manipulation score alongside the reasons and file clues. Confidence labels describe score ranges, not a validated accuracy rate. A high score is a reason for further review; a low score does not certify authenticity. An unavailable score is not a low score.

What are the limits of this image check?

False positives and false negatives are possible, especially after resizing, compression, retouching, or changes in image generators. FauxGuard does not verify C2PA signatures or SynthID watermarks, identify a person, or provide a local edit heat map. It accepts JPG, PNG, and WebP still images, not video or audio. Results are informational, not forensic or legal findings.