High-fidelity fashion
imagery at scale.
An SDXL-based image generation pipeline for fashion and editorial photography. Identity-consistent output across outfit variations using LoRA training and PhotoMaker encoding. Runs entirely on-device — no cloud API required.
- Platforms
- macOS, Docker
- Category
- AI
- Built for
- Fashion brands, e-commerce, content creators
- Status
- Live
What it does.
Identity LoRA Training
Train a custom identity model from 12–30 reference images. Output is consistent across all generated variants.
SDXL + 8K Upscaling
Two-stage pipeline: SDXL base generation at 1344×768 then ESRGAN 4× upscale to 8K.
PhotoMaker Encoding
Identity features extracted and encoded per-image using PhotoMaker and IP-Adapter FaceID.
50+ Outfit Library
Curated haute couture, tailoring, and editorial outfit prompts with dynamic parameter injection.
On-Device Only
Runs on Apple Silicon via PyTorch MPS. No cloud API calls, no data leaves the machine.
- Python 3.10
- SDXL
- ComfyUI
- PyTorch MPS
- LoRA
- Docker
Questions about Studio?
Ask about availability, licensing, or how it would fit what you're building.