Skip to content
AtheronLABS

You're visiting from the United States. Prices are shown in US dollars. Not right?

labs@atheron:~/industries/music-film/ai$ train agent --grounded

Metadata clean-up AI

Suggested fixes for your catalogue, approved by your team.

Catalogue metadata collects small errors over years: a name spelled three ways, a missing credit, a code on the wrong edition. An open-source model can find and suggest fixes for these at scale, on servers you control, while your team keeps the final word on every change.

// what matters here

What is different in this industry.

01

Suggestions, not changes

The model proposes; your team approves or rejects each fix.

02

Your catalogue stays private

The model runs on servers you control, and nothing is sent to an outside AI provider.

03

Measured on your catalogue

Scored against fixes your team has already made, before every change to the model.

04

Credits handled with care

Changes to creator credits always go to a person, never applied automatically.

// example projects

Priced examples for this service.

Each one is hypothetical, labelled as such, and priced live from our rate card. Open any of them in the estimator and make it yours.

Example projectAPIAIOpen-source model

Catalogue metadata and distribution feeds

Hypothetical. Not a client, not a result.

// The problem

A label's catalogue metadata lives in spreadsheets, and titles, credits, codes and release dates differ between what it sends to each distributor and what appears in the stores.

// What we would build

An integration hub that keeps one catalogue record and sends it to each distributor in the feed format they accept, such as DDEX for music or ONIX for books, with an open-source model that suggests fixes for missing credits, inconsistent names and mismatched codes for your team to approve.

// What is in it

  • One catalogue record for titles, credits, codes and dates
  • Feeds to each distributor in the format it accepts, such as DDEX or ONIX
  • Suggested fixes for credits, names and codes, approved by your team
  • Changes sent to every distributor at once, with a log of each delivery
  • A model run on servers you control, so the catalogue stays private

// Stack

  • Node.js
  • PostgreSQL
  • DDEX and ONIX feeds
  • Open-source model on GPU
  • DigitalOcean Toronto

// estimate

Build
≈ US$108,000 to US$165,000, delivered within 22 weeksCAD 153,300 to 234,400
Hosting
≈ US$4,360 a monthCAD 6,210 a month
Support
≈ US$2,500 a monthCAD 3,565 a month

Prices in your currency are estimates from today's Bank of Canada rate. All invoicing is in CAD or USD.

AI route
Open-source model
Model running cost
≈ US$4,190 a monthCAD 5,960 a month

Timeline by milestone

Specification and evaluation plan
2.9 to 3.4 weeks
API contract agreed
2.4 to 3.4 weeks
Working pilot
2.9 to 4.9 weeks
Core resources built
1.9 to 3.4 weeks
Security, gateway and connections
1.4 to 2.4 weeks
Docs, SDKs and load tests
1.4 to 1.9 weeks
Production
0.9 to 1.4 weeks
Launch and first callers
0.9 to 1.4 weeks

Outside our hands, and added to the calendar

API access to each outside system
1 to 4 weeks

How it is paid

Deposit 20%
CAD 30,600 to 46,820
Specification and evaluation plan 6.8%
CAD 10,404.00 to 15,918.80
API contract agreed 4.4%
CAD 6,732.00 to 10,300.40
Working pilot 20.4%
CAD 31,212.00 to 47,756.40
Core resources built 13.4%
CAD 20,502.00 to 31,369.40
Security, gateway and connections 8.9%
CAD 13,617.00 to 20,834.90
Docs, SDKs and load tests 4.4%
CAD 6,732.00 to 10,300.40
Production 6.8%
CAD 10,404.00 to 15,918.80
Launch and first callers 4.9%
CAD 7,497.00 to 11,470.90
Holdback, 30 days after launch (10%)
CAD 15,300 to 23,410
GPU time for training and testing, at cost
CAD 300. Billed up front, at cost, outside the milestones

// questions

Questions we get about this.

Does the metadata model train on our artists' work?

No. It reads catalogue metadata only, never the recordings, films or manuscripts.

How accurate are the suggested fixes?

We measure it on fixes your team has already made and show you the results before it goes live.

Can it fill in missing credits on its own?

No. It flags gaps and suggests where the answer may be; a person confirms every credit.

// next

Not quite your project?

Tell us what you have in mind. We will come back to you with a range and the questions that would narrow it. Or book a call and talk it through.

Open-source AI for catalogue metadata clean-up at labels and publishers | Atheron Network Labs