Technology & product direction
A more complete picture of fit.
We are developing Fitality around three connected ideas: personal body profiles, garment understanding and transparent evaluation. The public demo illustrates the direction, not a finished AI service.
01Intended architecture
The intended pipeline
Planned · not a live service
- 01
Input quality checks
Checks that photos are usable: framing, pose, lighting and clothing.
- 02
Body-profile estimation
Estimates a personal body profile from front and side photos and optional measurements.
- 03
Garment information
Combines garment images with available measurements and construction details.
- 04
Try-on generation
Generates a visual preview of the garment on the profile.
- 05
Evaluation and uncertainty
Compares body and garment information and reports limitations alongside results.
Reasoning and orchestration
We plan to use Claude for reasoning, structured information extraction and workflow orchestration alongside specialized computer-vision and generative models. The public sample demo does not call a live AI inference service.
02Our current stage
Starting with measurement validation.
Our first focus is comparing body-measurement estimates from two-view photos with real measured data. Personalized try-on, fit analysis and retailer integrations are later steps in the product direction.
What we still need to validate
Photos vary, measurements can be uncertain and fabrics behave differently. We are starting with measurement validation before the full personalized try-on and fit-analysis layers. No accuracy guarantee or launch date is being announced.
03The public demo
What the demo shows — and what it doesn't.
It does
- Shows curated sample images of two synthetic models in three looks and three views
- Aligns a reference outfit with a selected look for comparison
- Illustrates circumference differences with fictional measurements in Fit Lab
It does not
- Upload, analyse or store photos of you
- Run a live AI model or generate images on request
- Recommend a size, predict fit or measure your body
04Uncertainty
Why fit is uncertain.
- Photos vary
- Distance, lens, posture, lighting and clothing all change what a photo shows about proportions.
- Measurements are uncertain
- Even tape measurements differ between people and attempts. Estimates should carry a range, not a single confident number.
- Fabrics behave differently
- Stretch, drape, weight and construction change how two garments with the same measurements feel and look.
- A preview is not proof
- A convincing image can still hide a poor fit. Visual previews need to be paired with honest measurement context.
05Future data handling
Principles we are designing toward.
These are design principles for future features, not a description of a live product. Today's website collects only what the privacy notice describes.
- 01Ask only for the inputs a feature needs, and explain why before asking.
- 02Treat body photos and measurements as sensitive, with clear retention and deletion choices.
- 03Keep people in control: no hidden reuse of personal images for other purposes.
- 04Publish how processing works before launching photo-based features.