Content Authenticity API
Is this media real, AI-generated, or edited?

How we compare

The incumbents are enterprise deepfake platforms sold through a sales motion and an annual contract (Reality Defender, Sensity), or volume-priced moderation APIs that treat AI-generated content as one label among many (Hive). The hyperscalers don't ship a dedicated authenticity or provenance API at all. Brainiall is a flat per-asset REST call: send the media, get a verdict with the signals attached — including authoritative C2PA Content Credentials extraction — self-serve from the first call.

Prices are list-price approximations for orientation, not quotes — most authenticity vendors price by negotiation. Pixel-only deepfake detection is an unsolved, adversarial problem industry-wide; Brainiall is honest about that — see "How the score works" below. Always check each vendor's current terms.

Pricing

One per-asset price, whatever the media type. The free tier is enough to wire authenticity checks into an upload or onboarding flow and see real verdicts.

PAYG: $0.01 / analyzed asset (Brainiall Authenticity engine), any media type. No minimum spend, no contract — the same single API key and usage-based billing as the rest of the catalog.

Four endpoints

# Submit media as raw bytes, a multipart file upload, or {"data": "<base64>"}.

# 1. Image — AI-generated / manipulated detection
POST https://api.brainiall.com/v1/authenticity/image
  -> {"media_type": "image", "ai_generated_likelihood": 0.98, "verdict": "ai_generated",
      "signals": [
        {"signal": "content_credentials_ai", "tier": "provenance", "weight": 4.2,
         "direction": "synthetic",
         "detail": "Embedded C2PA Content Credentials declare this as AI-generated media."},
        {"signal": "exif_ai_software", "tier": "provenance", "weight": 3.3,
         "direction": "synthetic",
         "detail": "The EXIF Software tag references a known AI image generator."}],
      "engine": "Brainiall Authenticity engine"}

# 2. Provenance — extract embedded C2PA Content Credentials
POST https://api.brainiall.com/v1/authenticity/provenance
  -> {"has_content_credentials": true, "is_ai_generated": true,
      "claim_generator": "Midjourney/6.0", "signature_present": true,
      "assertions": ["actions", "ingredient (source asset)"],
      "summary": "Content Credentials are present and declare AI-generated origin."}

# 3. Audio — synthetic / cloned-speech detection
POST https://api.brainiall.com/v1/authenticity/audio
  -> {"media_type": "audio", "synthetic_likelihood": 0.55, "verdict": "uncertain",
      "signals": [{"signal": "bandwidth_ceiling", "tier": "forensic", "weight": 0.55, ...}]}

# 4. Video — deepfake detection (frame sampling + temporal consistency)
POST https://api.brainiall.com/v1/authenticity/video
  -> {"media_type": "video", "deepfake_likelihood": 0.71, "verdict": "likely_ai_generated",
      "frames_analyzed": 8, "per_frame_likelihood": [0.55, 0.74, ...], "signals": [...]}

Every response is fully explainable: signals lists exactly what was found and how much it moved the score. Each signal carries a tier — provenance (deterministic: embedded C2PA / metadata, authoritative when present) or forensic (heuristic indicator). No GPU, no opaque classifier.

How the score works

Honesty first: pixel-only deepfake detection is an adversarial, unsolved problem. So the engine is built around what it can verify deterministically.

• Provenance signals are authoritative. Embedded C2PA Content Credentials, XMP/IPTC DigitalSourceType markers and EXIF generator tags are deterministic — when present, they decide the verdict. C2PA is the emerging industry standard, embedded by DALL·E, Adobe Firefly, and a growing list of cameras and tools.
• Forensic signals are honest indicators. Fourier-domain artifacts, error-level analysis, spectral forensics and dimension fingerprints are corroborating heuristics — never decisive on their own.
• The engine never bluffs. Without a provenance signal it returns at most uncertain on a single forensic indicator, or likely_ai_generated on several — it only asserts a definitive ai_generated verdict when it has authoritative provenance. An ordinary recompressed photo will read uncertain, not a false positive.
• You see the reasoning. Every verdict ships with the full signal list, so a human reviewer can read why — and decide.

What it's for

• KYC & onboarding integrity: flag AI-generated or manipulated ID photos and selfies before they pass identity verification.
• Trust & safety: screen user-uploaded images, video and audio for synthetic and deepfake content at upload time.
• Newsroom, OSINT & provenance: extract and surface C2PA Content Credentials to verify where an image came from before it's published or cited.
• Insurance & claims: catch AI-generated or edited claim photos and supporting documents.
• Marketplace & listings: detect AI-generated product imagery and synthetic reviews-with-media.
• One bill, one key: same Brainiall API key and usage-based billing as the rest of the catalog — no separate authenticity vendor to procure.

Press kit & resources

What reviewers, integrators and procurement teams typically ask for.

More specialty APIs

Same single API key, same usage-based pricing, different problem solved.

Identity VerificationFace match, liveness & deepfake checks

Content ModerationText & image policy screening

Document AIOCR & structured extraction from documents

Vision LabelsObject & scene tagging for images