Better Together Trace by Soundverse · on Azure Marketplace
Audit-grade audio attribution for the AI era.
Trace is built for rights-holders who need evidence, leverage, and enforcement — not dashboards that stop at detection. Where traditional fingerprinting tools tell you that two tracks sound similar, Trace tells you which track, which stem, which section, and at what level of similarity, with audit-grade evidence packaged for legal review and takedown. Procurable through Azure Marketplace. Aligned with Microsoft's AI governance posture.
Marketplace billing retires against MACC. Compute, storage, and search inherit Microsoft's procurement, security, and DPIA frameworks.
- 2Reference corpora — internal training data and partner catalogs
- 9Deep search use cases — from cover detection to voice timbre to motif isolation
- 4Depth tiers — track, section, stem, and motif-level inspection
- N↔NMany-to-many mode for fleet-wide deduplication and catalog hygiene
Untitled · Submission #2026-1140
Pre-release screening · 2:47 · 142 BPM · A minor
Search mode · 1:N · Stem-level depth · Motif scan ON
“Midnight Echoes” — registered 2024
Partner catalog · ID 7714 · 142 BPM · A minor
Result: High similarity · Section 1:32–2:08 flagged · evidence packet ready
Built for legal teams who need to act on findings — not just look at them.
Why Trace is different
Most fingerprinting tools stop at “these tracks sound similar.” Trace tells you which stem, which section, and at what similarity — with evidence formatted for takedown.
Stem & section depth
Track-, section-, stem-, and motif-level inspection. Catch a vocal motif borrowed across the bridge of two otherwise-different productions. Pay for precision where it matters; use cheap track-level for triage.
See depth tiersTwo corpora, two roles
Internal training data (memorization audits) and curated partner catalogs (operations and royalties). Explicit corpus separation between audits and catalog operations.
See corporaEvidence, not dashboards
Every match comes with a packaged evidence bundle: similarity scores per stem and per section, time-aligned excerpts, and the audit trail of which corpora were searched. Drop into a takedown filing or a legal review.
View sample evidenceProcurable on Azure
Listed on Azure Marketplace with private-offer pricing for catalog-scale workloads. Compute, storage, and search retire against MACC. Trace inherits Azure's security, compliance, and audit posture out of the box.
Read Marketplace listingA specific tool for every shape of similarity.
Nine deep-search use cases
Trace exposes nine distinct search behaviors. Choose the one that matches your workflow — or run several in parallel for combined evidence.
- 01
Comprehensive Deep Search
End-to-end unified search across melody, lyrics, samples, timbre, and genre. The default for triage.
- 02
Genre / Style Deep Search
Detects stylistic and genre-level overlap to flag derivative or misclassified works.
- 03
Cover-Song Detection
Tempo- and key-invariant matching. Catches the same composition under different production choices.
- 04
Remix / Derivative Detection
Structural and motif-level reuse across edits, mashups, and DJ tools.
- 05
Lyrics Similarity
Near-duplicate and paraphrase detection across sung lyrics, with cross-language coverage.
- 06
Motif Detection
Isolation and identification of repeated phrases from vocal and instrumental tracks.
- 07
Melody Similarity
Predominant-melody extraction with contour matching, robust to time-warping and ornamentation.
- 08
Sample Search
Section-, phrase- and stem-level matching. Finds re-used samples across catalogs.
- 09
Voice Timbre Search
Voice and tone matching on the vocal stem. Flags a reused vocal recording, not the same singer in a new performance.
Pay for precision where it matters; use cheap depth for triage.
Depth tiers
Trace exposes four depth tiers. Most workflows triage at track-level and escalate to stem or motif-level only when a flag warrants it.
| Depth | What it inspects | Use profile | Relative cost |
|---|---|---|---|
| Track-level | Whole-song embeddings | Good recall, lower precision on short borrowings. Use for first-pass triage across catalogs. | 1× baseline |
| Section-level | Intro / verse / chorus / bridge / outro | Catches localized reuse — chorus borrowed across otherwise-different productions. | ≈1× per layer |
| Stem-level | Vocals, drums, bass, other | Big lift on mix-masked reuse. Critical for vocal motifs and reused vocal recordings. | ≈1× + stem split |
| Motif-level | Short phrases (1–8 bars) | Highest precision on melodic borrowing. Reserved for high-stakes legal review. | per song pair |
Two corpora. Two operational roles. Explicit separation.
Reference corpora
The corpus you search against determines what conclusions you can draw. Trace separates training-set audits from catalog operations — by design.
| Corpus | Description | Operational role |
|---|---|---|
| A · Internal training data (closed) | Used to detect memorization in the model itself. Never used for public takedowns. | Compliance and AI governance. Highest value for model audits. |
| B · Partner catalog (consented) | Rights-holders onboard catalogs via the Partner Program or via secure storage URL. | Business operations and royalty payouts. Clean metadata, consented use, attribution hooks. |
One-to-one for legal review. One-to-many for screening. Many-to-many for catalog hygiene.
Search modes
Three search topologies cover the workflows rights-holders actually run.
One-to-One
Compare a candidate against a single suspected reference. Fastest path for triage and internal legal review of a specific allegation.
One-to-Many
Compare a candidate against a label catalog. The workhorse mode for pre-release screening and partner safety.
Many-to-Many
Continuous fleet-wide deduplication, remix discovery, and cover detection across rolling batches. For catalog hygiene and market monitoring.
A generic view of how Trace is procured and operated on Azure.
Reference architecture
Audio in, evidence out. Trace runs as a managed service on Azure-deployable infrastructure, with Marketplace handling procurement and Azure handling identity, storage, and observability. The internal scoring engine is intentionally abstracted here — implementation details are covered under partner NDA.
Microsoft × Soundverse · Joint GTM
Co-positioned as the third-party attribution layer for Microsoft's AI governance posture.
As Microsoft's own audio models advance, Trace operates as an independent, audit-grade attribution layer — procurable through Marketplace with the same governance rigor Microsoft customers expect. For label legal teams, PROs, and DSPs, that means one vendor for both creation (Soundverse API + Studio) and rights protection (Trace), all on Azure.
- ✦ Marketplace ISV
- ✦ AI governance partner
- ✦ Independent attribution
- ✦ MACC-eligible
Joint motions in market
- Major-label rights protection — co-pitched into UMG, Sony Music, and Warner with Azure compute commitment.
- PRO & CMO attribution — ASCAP, BMI, PRS, GEMA, JASRAC: statutory-grade audit trail for derivative-work distribution.
- DSP pre-ingest screening — Spotify, Apple Music, YouTube, TikTok facing AI-upload volume; Trace flags or attributes at ingest.
- Catalog acquisition diligence — Hipgnosis, Round Hill, Primary Wave: originality due diligence before purchase.
Anyone whose business model depends on knowing whether audio is original, derivative, or memorized.
Built for these buyers
Nine distinct buyer profiles — each with their own decision-maker, procurement cycle, and outcome metric.
Music labels
Pre-release screening, catalog hygiene, AI-leakage audits, and remix / cover discovery across major and independent catalogs.
PROs & CMOs
Royalty distribution accuracy, derivative-work attribution, and AI-music registration. Statutory-grade audit trails.
Music publishers
Mechanical and sync attribution, sample-clearance workflows, and catalog enforcement at the composition level.
DSPs
Pre-ingest screening to block infringing uploads, attribute legitimate derivatives, and reduce takedown burden.
UGC platforms
Content ID-style attribution for AI-generated audio, with stem-level granularity. Reject, attribute, or allow with metadata.
Catalog acquirers
Due diligence on catalog originality before acquisition. Uninsurable risk in an AI-generated era — Trace is the diligence layer.
AI governance teams
Internal memorization audits and pre-deployment regression testing for any team shipping generative-audio models.
Media production
Internal sync screening, AI-output audit, and clearance documentation for film, TV, ad, and streaming production.
IP & entertainment law
Evidence packaging for litigation and takedown actions. Audit-grade reports formatted for legal review.
One audio AI platform. Four enterprise products.
More from Soundverse on Azure
Trace is the trust layer; three other offerings serve different buyers and motions on Azure Marketplace.
Programmatic audio AI for ISVs and developers.
RESTful and streaming endpoints for generation, modification, and intelligence. Code samples in Python and JavaScript. API keys. 99.9% SLA. Five commercial-rights tiers including full-ownership.
Explore the APIFigma for audio + video, with AI as native instruments.
Browser-based real-time multi-track studio. Up to 50 collaborators per project on paid plans. Full generative-AI tool surface inside the timeline. Video import, video generation, multi-resolution export.
Explore the StudioMusic creation and Teaching Agents for schools.
Bulk institutional licensing. Classroom workflows. The Teaching Agent — a personalized AI replica of any educator, trained on their style, available to students 24/7.
Explore EducationShip audio knowing the rights story holds up to audit.
Run a pilot scan against your catalog, or talk to the Soundverse partner team about a private offer for catalog-scale workloads.
Microsoft and Azure are trademarks of Microsoft Corporation.