Loading two-layer diagram...

Honest comparison

SynthID and OmniPulse answer different questions.

SynthID answers a different question. Both matter.

DimensionSynthID (Google DeepMind)OmniPulse
Who owns the recordThe providerThe creator, on a decentralized ledger
What it verifiesWhether content was AI-generatedWhether content is yours, and under what license
How verification worksThrough the provider's serviceDeterministic mathematics anyone can run

Loading embed simulator...

Loading pipeline simulator...

Verdicts

Three outcomes. One unambiguous number.

Every comparison returns one of three verdicts. The thresholds below are provisional: they will be recalibrated on real-world data before the platform exits private beta.

Exact

The fingerprints match within the exact-match threshold. The query file is a bit-identical or perceptually identical copy of the registered original.

Perceptual

The Sliced-Wasserstein distance falls between the exact and perceptual thresholds. The content is likely a derivative, pitch-shifted copy, or re-encode.

Miss

The distance exceeds the perceptual threshold. No match found in the catalogue for this query.

Threshold note: Exact and Perceptual thresholds are set conservatively and have not been calibrated against a production corpus. Treat all Perceptual verdicts as candidates for human review until calibration data is published.

Signed attestations

Ed25519. Three public fields. No account required.

Every match verdict is accompanied by a signed attestation. The signature is verifiable by anyone who holds the three public fields below, with no call back to OmniPulse and no account. The signing key is the organization's own key, not ours.

asset_id

The SHA3-256 hash of the fingerprint vector at registration time.

owner_key

The Ed25519 public key of the registering organization.

timestamp

Unix timestamp of registration, included in the signed payload.

Modular commercialisation

Four tiers, one fingerprint plane

Phase I

Phase I: Institutional Research

Standalone DSP wheels for desktop labs and HPC clusters. C++/CUDA Wavelet Scattering primitives, no orchestration required. AGPL-3.0 + Commercial.

Scope:Passive audio only.

pip install omni-wst-core

Phase II

Phase II: High-Velocity Concurrent Processing

The vector substrate. omni-hnsw + sliced-wasserstein Rust crates for fast nearest-neighbour over large-scale fingerprint catalogues. AGPL-3.0 + Commercial.

Scope:Passive across all modalities, small catalogue.

cargo add vector-index sliced-wasserstein

Phase III

Phase III: Autonomous Agentic Control

Drop-in MCP server. omnipulse-agent parses operator requests and routes them to the Rust orchestrator over line-delimited JSON-RPC 2.0. AGPL-3.0 + Commercial.

Scope:Active embed for images and video, full catalogue.

pip install omnipulse-agent

Phase ENTERPRISE

Enterprise SaaS Pipeline

Commercial

Cloud-native multi-tenant deployment. Kubernetes control plane, GPU autoscaling driven by Prometheus queue depth, ed25519-signed license tokens, IPFS pinning.

Scope:Dedicated instance, versioned verification procedures.