Skip to navigation
API ReferenceDeepfake Detection

Create deepfake detection

Analyze audio, image, and video for deepfake detection.

Supply media via one of three intake methods:

  • Direct file upload — multipart/form-data with the file attached as file. Files must be 150 MB or smaller and use one of the supported audio/video/image extensions. For larger files, use the secure upload flow.
  • Public URL — application/json with a url field. The API fetches the URL itself.
  • Secure upload token — application/json with a media_token field obtained from POST /secure_uploads.

Exactly one of file, url, or media_token must be provided per request.

Set detect_watermark=true to run Resemble watermark detection and SynthID alongside the deepfake analysis. This option supports single audio, image, and video requests only. It adds one Watermark detection charge, applies a 200 MB source limit for audio, image, and video, and causes Prefer: wait and callbacks to wait for the watermark analysis to reach a terminal state. A watermark failure is returned in the nested watermark object and does not change the deepfake verdict.

Authentication

AuthorizationBearer

Request

This endpoint expects a multipart form containing a file.
filefileRequired

The media file to analyze (audio, video, or image). Must be 150 MB or smaller.

callback_urlstringOptionalformat: "uri"
POST destination for the detection result.
wait_for_intelligencebooleanOptional

Controls callback timing when callback_url and intelligence=true are provided. True waits without a forced timeout until Detect and Intelligence are both finished. False sends after Detect finishes without waiting for Intelligence. When omitted, the callback waits up to 10 minutes for Intelligence before sending the result state available at that time.

visualizebooleanOptional
Generate visualization artifacts
frame_lengthintegerOptional1-4Defaults to 4

Window size in seconds (audio/video). We recommend 4 for audio.

start_regiondoubleOptional

Start of segment to analyze (seconds)

end_regiondoubleOptional

End of segment to analyze (seconds)

max_video_secsdoubleOptional
Cap processed duration
modalityenumOptionalDefaults to all

Video only. Selects which detection components run for a video file: audio (audio track only), video (visual frames only), or all (both — the default). Ignored for audio and image inputs. Invalid values are rejected with HTTP 400.

Allowed values:
face_onlybooleanOptionalDefaults to false

Focus visual video detection on faces by masking non-face regions. Effective only for video inputs with modality=all or modality=video. Audio, image, and modality=audio video requests are accepted but use an effective value of false.

intelligencebooleanOptionalDefaults to false
Run multimodal intelligence analysis on the media
detect_watermarkbooleanOptionalDefaults to false
Run Resemble watermark detection and SynthID. Supported for single audio, image, and video requests. Adds the Watermark detection charge and applies a source limit of 200 MB for audio, image, and video.
infer_from_intelligencebooleanOptionalDefaults to false

Opt in to let a strong intelligence finding escalate an otherwise non-fake verdict to “Likely Fake”. Has no effect unless intelligence is also true.

audio_source_tracingbooleanOptionalDefaults to false
Enable audio source tracing to identify synthetic audio origin
signalbooleanOptionalDefaults to false

Run Signal content analysis after detection completes, classifying the media against fraud and abuse categories. Requires Signal access on your plan and cannot be combined with zero_retention_mode. Not supported for batch detections.

use_ood_detectorbooleanOptionalDefaults to false

Enable out-of-distribution detection

zero_retention_modebooleanOptionalDefaults to false
Enable Zero Retention Mode to automatically delete submitted media after detection completes.

Response

Detection job created
successbooleanOptional
itemobjectOptional

Errors

400
Bad Request Error
402
Payment Required Error
403
Forbidden Error