Recognition that never calls home

Know what’s playing.
Anywhere.

Build and ship private audio-recognition catalogs. Fast enough for live detection. Compact enough for mobile and edge devices. Fully offline at match time.

<50 msCold start
100%Offline matching
10M+Fingerprints
Live recognition
MATCH FOUND · 98.7%Summer launch · 30s00:12.480 offset
LIVE
catalog_eu_v14.acrf On-device
6 ms reference match latency
Built forBroadcastersRetail mediaMeasurement teamsConnected devices
Offline detectionAvailable now
Available
Online detectionCloud-assisted recognition
On request
Live channel detectionContinuous channel monitoring
Beta
From audio to recognitionREADY FOR THE EDGE
01Add mediaAudio or video
02PrepareHandled for you
03BuildYour catalog
04RecognizeOn the device
PRIVATE Your mediaPORTABLE One catalogOFFLINE Fast answers
One recognition platform

From source creative to an answer at the edge.

Signals gives teams a controlled way to build, version, distribute, and operate private audio-recognition catalogs.

Shared fingerprint engine

Database generation and device matching use the exact same Go implementation, keeping every hash aligned.

CONSISTENT BY DESIGN

Compact in-memory catalogs

Hash-grouped posting lists are loaded fully into memory. No SQL queries, service calls, or network lookup at match time.

FORMAT V3 · ACRF

Immediate recognition

Short microphone buffers vote on sound and offset together, producing useful results without waiting for an entire clip.

SOUND ID + OFFSET
Recognition in the real world

Designed for the places connectivity cannot be assumed.

Keep raw listening data local while giving product and operations teams a managed catalog workflow.

ConnectionOptional
Store
Mobile
Broadcast
Build an offline project

Broadcast monitoring

Recognize ads and creative variants from live television or radio captures without sending recordings to a cloud matcher.

Retail media verification

Confirm what played on in-store screens and speakers—even across locations with unreliable connectivity.

Mobile experiences

Embed private recognition inside Android apps for second-screen experiences, attribution, and interactive campaigns.

Connected devices

Run compact catalogs on edge hardware where latency, bandwidth, and data ownership matter.

A calm operational workflow

Three steps from media to recognition.

Add your creatives, keep everything organized, and publish a recognition-ready catalog when you are ready.

01

Add your media

Upload the audio and video your product needs to recognize. Signals prepares each creative for you.

02

Organize your catalog

Keep creatives, tags, and projects together in one workspace your whole team can understand.

03

Publish when ready

Create a release, add it to your product, and keep recognition content current as campaigns change.

Ready to releaseOne clear catalog package
Built to run anywhereAdd it to your product
Useful answersKnow what played and when
Ready when you are

Turn your first creative into a recognition-ready catalog.

Start your catalog
Android deviceLive audio input
short buffer
Native matcherIn-memory vote index
local result
Creative #42Offset 00:12.480
At match time

No cloud lookup. No SQL. No audio egress.

The prebuilt fingerprint database stays in memory beside the native matcher. Captures never need to leave the device, and results arrive without a network round trip.

≈6.5 bytesper fingerprint posting65,535sounds per catalog
Control plane security

Private matching, governed operations.

The runtime is offline. The catalog workflow still has the identity, tenancy, and automation controls a real team needs.

Your organizationTenant-isolated projects and mediaPROTECTED
Role-based accessOwner · Admin · Member · Viewer
Scoped API keysOrganization or project · Read or write
Private object storageNormalized WAVs and versioned catalogs

Hashed credentials

Argon2id passwords and SHA-256 session, reset, and API-key secrets.

Automation without sessions

Upload media, rebuild catalogs, and download outputs using scoped bearer keys.

Server-side authorization

Every tenant route resolves membership or API-key scope before touching project data.

Built for automation

Your catalog pipeline belongs in CI, too.

Use scoped write keys to add creatives and rebuild. Use read keys in controlled delivery jobs that fetch the latest catalog.

Create your workspace
# upload a creative
curl -H "Authorization: Bearer $SIGNALS_KEY" \
  -F "[email protected]" \
  /api/orgs/acme/projects/tv-ads/media

# build the offline catalog
curl -X POST -H "Authorization: Bearer $SIGNALS_KEY" \
  /api/orgs/acme/projects/tv-ads/offline-db/rebuild
Simple pricing

Choose how you want to run Signals.

Own the stack with a one-time license, start managed by the audio hour, or design a larger deployment with us.

Self-hostedOwn your deployment
$9,999$4,999
one-time, lifetime license

For teams that want to operate the catalog pipeline inside their own infrastructure.

Contact us
  • Signals dashboard
  • Offline catalog builder
  • Native Android matcher
  • Organization and project controls
EnterpriseFor larger deployments
Custombuilt around your scale

For large catalogs, high-volume media operations, or tailored integration requirements.

Request a quote
  • Volume pricing
  • Architecture guidance
  • Integration support
  • Deployment planning
Technical FAQ

The questions engineering teams ask first.

Signals separates catalog operations from runtime recognition, so each side stays simple and accountable.

01Does recognition require an internet connection?+

No. The native matcher and prebuilt ACRF catalog run entirely on the device. The dashboard is used to manage creatives and build distributable catalog files.

02What is Offline detection?+

Offline detection is available now. Your media database and matcher run locally on a mobile, edge, or Linux device, so recognition continues without a network connection.

03Is Online detection available?+

Online detection is available on request for products that prefer cloud-assisted recognition, centralized catalog access, or server-managed matching.

04What is Live Channel detection?+

Live Channel detection is currently in beta. It is designed for continuous monitoring of broadcast, streaming, or other always-on audio channels.

05What does a match return?+

The engine returns the recognized sound ID and the estimated offset where the captured audio begins inside that creative.

06Can the fingerprint pipeline be changed later?+

Yes, but database generation and device matching must use identical parameters. Any algorithm or pipeline change requires re-fingerprinting and redistributing the catalog.

07How is catalog access controlled?+

Organizations support owner, admin, member, and viewer roles. Scoped read/write API keys can automate uploads, builds, and downloads without sharing user sessions.

Talk to the team

Have a recognition problem worth solving?

Tell us what you need to recognize, where matching will run, and the scale of your catalog. We’ll help you map the shortest path to a working deployment.

Built for technical conversationsArchitecture, evaluation, integration, and deployment.
Contact usWe’ll get back to you within 1 business day.
Start with one catalog

Give your product ears
without giving up control.

Create a private recognition workspace, add your creatives, and ship an offline database built for the edge.

Start building