A multimodal deepfake detection API that accepts a file or URL and returns an Authentic / Likely Synthetic / Inconclusive verdict with a 0–100 TrustScore and confidence value. It covers three modalities: audio (detecting ElevenLabs, PlayHT, Resemble, OpenAI, Murf, and WellSaid output, plus cloning and spliced audio), images (Midjourney, DALL·E, Stable Diffusion, Flux, Firefly, Imagen, Ideogram, plus retouching and face swaps), and video (face swaps, lip-sync manipulation, fully generated video, with flagged segments returned as timestamps). Engineering details include file deletion after 60 seconds unless retain=true keeps it for 30 days for audit, official SDKs in eight languages (JavaScript/TypeScript, Python, Go, Ruby, PHP, Java, .NET, Rust), webhook-based async jobs, batch processing, a real-time streaming option for call-centre use, a 99.9% availability SLO across multi-region active-active infrastructure, and SOC 2 marked as “In Progress” rather than completed. Pricing: Free (50 checks/month, no card, 2-minute media limit), Starter $49/month or $39/month annual (1,000 checks, 10-minute limit, API access, PDF/CSV export), Business $199/month or $159/month annual (5,000 checks), Enterprise $599/month or $479/month annual (20,000 checks). Newsrooms and non-profits get 50% off Starter.
Pros & Cons:
Pros: Covers audio, image, and video in one API rather than requiring separate vendors per modality; returns a TrustScore plus flagged video segments with timestamps, which is more actionable than a binary verdict; eight official SDKs and webhook-based batch processing lower integration effort; files are deleted after 60 seconds by default, with an opt-in 30-day retention for audit purposes; a 99.9% availability SLO and multi-region active-active setup signal production intent; 50% discount for newsrooms and non-profits addresses a core verification audience.
Cons: No independent third-party validation surfaced—no academic benchmarks or independent reviews verify the claimed accuracy for media detection, unlike Pangram (NYT/University of Chicago) or Originality.ai (Epoch AI); Firefox extension text detection has documented failures—multiple 2026 one-star reviews report Claude, ChatGPT, and LLaMA text being labeled “likely human”, though this is the extension’s text feature, not the main media-detection API, and the two should not be conflated; SOC 2 is marked “In Progress”, not completed; the operating company name, team, and funding are not disclosed; third-party directory pricing data contains clear scraping errors (MEFMobile lists Starter at $4,939/month), so the official $49/month figure should be used instead.
Who is it for?:Newsrooms and fact-checking teams verifying audio, image, and video authenticity; platforms that need to screen uploaded media for synthetic content across multiple modalities; developers embedding deepfake detection into moderation or onboarding pipelines via SDK or webhook; call centres using the real-time streaming option to flag synthetic speech during live interactions.