What if your phone could grade vinyl
better than your eyes?
Camera-based surface analysis. LiDAR depth mapping. A machine learning model trained by the collector community. The first objective vinyl grading system that does not depend on human subjectivity.
In active development
Every vinyl transaction depends on condition. Mint, Near Mint, Very Good Plus. These grades determine whether a record is worth $5 or $500. And yet the entire system runs on one thing: the seller's eyes and honesty.
Two collectors can look at the same record and assign different grades. Lighting matters. Experience matters. Motivation matters. A seller grading their own inventory has every incentive to round up. A buyer receiving that record has no recourse except the return process.
The Goldmine standard has been the foundation of vinyl grading for decades. It is a good framework. But it was designed for human eyes, and human eyes are inconsistent. The standard itself is not the problem. The application of it is.
Groove Vision is an attempt to solve this. Not by replacing the Goldmine standard, but by applying it with a level of consistency that human vision cannot match.
The iPhone camera captures high-resolution images of the vinyl surface under controlled lighting conditions. The system analyzes micro-scratches, scuffs, hairlines, and pressing artifacts at the pixel level. Standard iPhone cameras provide enough resolution to detect surface imperfections invisible to the naked eye.
Scratch detection
Scuff mapping
Hairline analysis
Pressing defects
Surface wear index
Ring wear scoring
On iPhone Pro models, the LiDAR scanner adds a third dimension. It builds a 3D surface map of the vinyl, measuring groove depth, warp, and dish. A flat record and a warped record might look identical in a photo. LiDAR knows the difference. This layer is optional but significantly increases grading accuracy.
3D surface topology
Warp detection
Groove depth analysis
Dish measurement
Edge wear profiling
iPhone Pro enhanced
The neural network learns from every scan. Each time a collector scans a record and confirms or adjusts the suggested grade, the model gets smarter. Thousands of collectors scanning thousands of records, building the most comprehensive vinyl surface dataset ever assembled. The data is anonymous. The model is collective. The result improves for everyone.
Federated learning
On-device inference
Anonymous contributions
Goldmine-aligned output
Continuous improvement
Core ML framework
Groove Vision processes everything locally using Apple's Core ML and Vision frameworks. No images leave your phone. No server sees your records. The only data that leaves the device is anonymous, aggregated model feedback, and only if you opt in.
Input
Camera + LiDAR
Processing
Vision + Core ML
Output
Goldmine Grade
Feedback
Model Improves
Apple Vision Framework
Image analysis, feature detection, surface classification
Core ML
On-device inference, model updates via background transfer
ARKit + LiDAR
3D surface mesh, depth buffer, point cloud generation
Create ML
Model training pipeline, dataset management, validation
Swift + SwiftUI
Native scanning UI, real-time preview overlays
Open Source Dataset
Anonymized surface data, community-contributed, publicly available
A vinyl grading system only works if collectors trust it. And collectors will only trust it if they can see how it works, verify the data, and contribute to making it better. That is why Groove Vision is being built as an open research project.
The trained model will be open. The anonymized dataset will be open. The grading methodology will be documented and auditable. Any developer will be able to build on top of it. Any collector will be able to verify a grade.
Spinstack will be the first app to integrate Groove Vision, but it will not be the only one. If this works, it should be available everywhere vinyl is bought and sold. Discogs. eBay. Independent shops. The grading standard belongs to the community, not to any single app.
Trained Core ML model weights and architecture
Anonymized surface scan dataset with Goldmine grades
Training pipeline and validation methodology
API specification for third-party integrations
Grading methodology documentation
| Manual | Groove Vision | |
|---|---|---|
| Consistency | Varies by grader | Identical every time |
| Objectivity | Seller bias | No stake in outcome |
| Detection | Visible defects | Micro-level analysis |
| Warp detection | By feel | LiDAR measured |
| Speed | Minutes per record | Seconds |
| Improves over time | No | With every scan |
Groove Vision follows the same privacy principles as Spinstack. All image processing happens on-device using Apple's frameworks. No photos of your records are uploaded anywhere. No server ever sees your collection.
If you choose to contribute to the community dataset, only anonymized surface feature vectors are shared. These are mathematical representations of surface characteristics. They cannot be reverse-engineered into images. They contain no identifying information about you or your records.
Contributing is always opt-in. You can use Groove Vision to grade your records without ever sharing a single data point. The model still works. It just does not learn from your scans.
Phase 1: Foundation
Now
Camera capture pipeline. Initial surface feature extraction. Basic scratch and wear detection. LiDAR depth buffer integration on Pro models. First training dataset from internal scans.
Phase 2: Beta Scanning
2026
Release scanning UI in Spinstack as a beta feature. Collectors scan their records and confirm or adjust suggested grades. Every confirmation trains the model. Early results will be rough. That is expected and honest.
Phase 3: Community Dataset
2027
Open the anonymized dataset. Publish the model architecture and training pipeline. Invite third-party developers to build on top of it. Release the API specification for marketplace integrations.
Phase 4: Ecosystem
2028+
Groove Vision grades become a trusted standard in vinyl transactions. Integrated into marketplaces, auction houses, and record shops. The dataset grows continuously. The model accuracy approaches and potentially exceeds expert human graders.
Groove Vision starts inside Spinstack. Download the app, build your collection, and be among the first to scan when the beta arrives.
This is where it starts.