Private Catalogue Pilot // Managed evaluation · Powered by PYXIS-1
Test PYXIS-1 on the music that matters to your team.
A managed private pilot for labels, distributors and catalogue teams. Test sonic discovery on authorised material, with success criteria agreed up front.
MANAGED EVALUATION · PARTNER-SCOPED · NO PUBLIC SANDBOX
- Managed evaluation
- Representative catalogue sample
- Agreed scorecard
- Private partner index
- Joint findings review
- No public sandbox
- Scope
Agree the question, the sample and the success criteria.
- Ingest
Deliver audio + CSV manifest through managed SFTP.
- Index
Validation, PYXIS-1 processing, private index build.
- Test
Ranked results from your agreed reference set.
- Measure
Expert review against the agreed scorecard.
- Decide
Joint findings review: stop, extend or plan production.
01 // Fit
Built for catalogue teams with a real question to test
Can older tracks be surfaced from current references?
Are our catalogue relationships discoverable beyond metadata?
Where does our current catalogue search break down?
Can similarity fit inside our current product and journey?
Does the analysis produce useful structured data for briefs?
You must control or be authorised to supply the evaluation material. Submitting an inquiry starts a conversation about fit; not every request becomes a pilot right away.
- Authorised catalogue access
- Electronic-music repertoire
- Clear discovery or product question
- Representative sample available
- Expert reviewers available
- Willing to agree measures up front
- Very small or highly uniform sample
- No relevant comparison method
- No authorised audio delivery
- Unclear rights position
- Depends on roadmap features
02 // The sample
Bring a representative sample, not the full catalogue
We scope sample size and mix together, around your use case and the range of the catalogue.
MP3, WAV, FLAC or AIFF: material you control or are authorised to supply.
One row per track: ISRC, title, artist and filename, all required. BPM, duration and release details are optional.
A meaningful mix, with enough related recordings to test similarity properly.
Tracks your team already knows: the anchors the evaluation is scored against.
A small group of catalogue experts or selectors who will judge relevance.
The question the pilot should answer: discovery, catalogue or product.
Only where you want to evaluate product behaviour downstream.
no universal track count; the sample is scoped around the use case, not a number
03 // The method
One bounded process. Six clear stages.
Every stage states what Resomix does, what your team does, and what evidence it produces.
- Assesses fit against the qualification principles
- Proposes sample composition and evaluation method
- Drafts the scorecard for agreement
- Define the pilot question
- Nominate expert reviewers
- Confirm rights and sample availability
- Agreed scope document
- Agreed scorecard with targets
04 // The scorecard
Agree the scorecard before you see the results
No pre-filled percentages and no fabricated outcomes: targets are agreed in scoping, results are measured in the pilot.
Proportion of the first ten results your agreed reviewers accept as musically relevant.
agreed in scopingmeasured in pilotHow often catalogue experts consider the shortlist usable.
agreed in scopingmeasured in pilotWhether relevant older or low-exposure recordings surface.
agreed in scopingmeasured in pilotShare of the accepted sample searchable and represented in the index.
agreed in scopingmeasured in pilotHow often a valid reference fails to produce a useful shortlist.
agreed in scopingmeasured in pilotMedian and P95 response times, measured after the sample catalogue has been indexed. These are never guaranteed.
agreed in scopingmeasured in pilotMeasured separately, because the reference must first be processed before searching the partner catalogue.
agreed in scopingmeasured in pilotValidation success, rejected files, and time from accepted batch to evaluation-ready.
agreed in scopingmeasured in pilotsearch on indexed tracks and processing of uploads are timed separately, never merged into one "search speed" figure
05 // The output
A decision pack, not a sales demo
The pilot ends with evidence your team can act on: what worked, what didn't, where the limits are, and what production would involve. Final deliverables are agreed during scoping.
The evidence does not justify expansion. A clean, documented end.
A second bounded test, on another catalogue area or question.
Scope, ingestion cadence, API usage, service mode and terms, agreed together.
the pilot does not automatically convert into a contract · no public prices on this page
- Scope & criteria: the agreed questions, sample and scorecard
- Ingestion summary: validation results and the indexed sample
- Evaluation access: scoped API key or agreed evaluation interface
- Relevance evidence: ranked evaluations, scorecard, structured track-analysis examples scoped to the pilot use case
- Performance summary: query and ingestion measurements
- Catalogue observations: long-tail and metadata-review notes where relevant
- Limitations & open questions: stated plainly
- Production options: recommended next steps, reviewed jointly
06 // Control
Your sample stays inside the agreed pilot boundary
- Partner catalogues are logically isolated and restricted to authorised partner access.
- You keep ownership and commercial control of your catalogue throughout.
- Partner content is not used to train or fine-tune PYXIS-1. It is not a model, and it does not generate derivative audio.
- Public exposure is not part of the pilot.
- Your sample is processed within a partner-scoped catalogue boundary and restricted to authorised access. Detailed access, retention and deletion terms are agreed before ingestion. FINAL WORDING SUBJECT TO PILOT AGREEMENT
07 // Apply
Tell us what you want to explore
Tell us about your catalogue, API integration or potential pilot. Submitting this form doesn't create an account or automatically start a pilot.