What it does
Adapter Studio wraps a Low-Rank Adaptation (LoRA) fine-tuning pipeline for SDXL in a workbench that treats every training run as a first-class object: dataset, configuration, metrics, previews and the exported adapter live together. A few hundred mixed-format product shots become a clean training set without manual cropping, and the best checkpoint is picked by a fidelity score rather than a hunch.
What you get
- The full application as a Docker image, with the web UI, REST API and CLI.
- Reference training configurations for furniture, fashion and packaging photography.
- An inference service that swaps adapters at request time, so one base model serves every product line.
- Documentation, an onboarding session and a direct line to the engineers who built it.
Deployment and licensing
The one-time licence is perpetual, runs in your own infrastructure — a single GPU with 24 GB of memory is enough — and includes twelve months of updates. The subscription is hosted by us in an EU data centre, includes the compute, and can be cancelled monthly. Both include the source of the training pipeline so your team can extend it.
Proven in production
The product grew out of our work for a furniture brand: on a 312-image set the fidelity score rose from 0.61 to 0.88, and a full run finishes in under two hours. The case study linked above has the details.