Upload one photo of an outfit you like. FitCraft AI designs it for your shirt, pant and T‑shirt size, exports a tailor‑ready PDF your tailor can stitch from, and lets you try it on live through your camera.
Three sizes, one photo, one tailor‑ready PDF. Works on phone, tablet and laptop.
Built for Indian sizes · English and हिंदी · designed with tailors, not just algorithms
Planned on proven, open technology
Eight steps in the app, one promise behind them: a tailor pack accurate enough to cut cloth from.
Sign up with your phone number or email and confirm with a one‑time code.
Shirt (like 32), pant (like 40) and T‑shirt (S to XXL). Add height or fit preference for a closer match.
Your sizes live on a profile in your name. Edit them any time and re‑fit every saved design.
A model wearing it, or any screenshot of someone in an outfit you like.
The bot reads the garment, then rebuilds it from your measurements, not the model’s.
Keep every version, compare them, and re‑fit when your sizes change.
One PDF at your size: measurements, pattern pieces at 1:1, cutting layout and stitch notes.
Open your camera and see the saved outfit on you before a single cut is made.
1:1
print scale on every sheet
Checked with a 10 cm test square
±1mm
seam match between joined pieces
Automated check; a failed check blocks export
3
garments at launch
Shirt, pant and T‑shirt
<60sec
photo to saved design
v1 target, measured in the pilot
These are design targets for version 1, not measured results. They are validated with partner tailors before launch.
A photo cannot give exact measurements, so we do not pretend it does. The AI reads the style; your sizes and a pattern engine supply the numbers.
Pick the three sizes a user would enter. The pack engine turns them into the measurements a tailor works from. These are starter formulas for illustration; the shipped tables come from published Indian size surveys and the tailor pilot.
| Garment | Measurement | Inches | cm |
|---|
Open your camera and your saved, size‑matched outfit appears on your body in real time, in the spirit of the Anywear Chrome extension, but for the designs you made.
Pose tracking in the browser warps the garment onto your shoulders and torso. Fast, light, works on mid‑range phones.
Ships firstBetter sleeves, collar and occlusion when you move your arms. More build effort, so it is an upgrade.
Full tierA paid try‑on API renders one realistic photo. It is not live, and the open‑source models carry non‑commercial licences, so a commercial API is used.
Add‑onYour video is processed on your device and never uploaded. Only the outfit you save and your sizes are stored.
Requirements, architecture, a nine‑phase build plan and a cost model in rupees, all on this page.
Each of your nine steps becomes testable requirements, plus the quality bars that make it work on every phone, tablet and desktop.
Accept: a new user reaches the size screen in three screens or fewer.
Accept: impossible combinations are flagged; every save is versioned.
Accept: one action re‑fits all designs and tells the user what changed.
Accept: preview shown with a clear progress state; failures explain what to retake.
Accept: 100% of exported specs pass schema validation.
Accept: no design is lost on refresh or a dropped connection.
Accept: the 10 cm test square prints at 100 mm ±1 mm on a correctly set printer.
Accept: smooth on mid‑range Android and recent iPhones; verified with a network log.
Accept: pilot garments measured against the pack; results published internally.
| Table | Key fields |
|---|---|
| users | id, phone, email, birth_year, language, guardian_id, created_at |
| profiles | user_id, shirt_chest, pant_waist, tee_size, height_cm, fit, units, version |
| size_history | user_id, version, values (JSONB), changed_at |
| uploads | id, user_id, storage_key, status, moderation, expires_at |
| designs | id, user_id, upload_id, garment_spec (JSONB), preview_key, size_version, engine_version, status |
| packs | id, design_id, pdf_key, params_hash, pages, created_at |
| fit_feedback | pack_id, area, direction, amount_cm, source (user/tailor) |
| consents | user_id, purpose, notice_version, given_at, withdrawn_at, guardian_proof |
| usage_costs | design_id, model, tokens/images, cost_inr |
| Method | Path | Purpose |
|---|---|---|
| POST | /auth/otp | Request and verify a code |
| GET | /me/profile | Read sizes and settings |
| PUT | /me/profile | Update sizes (new version) |
| POST | /uploads | Signed URL for an image |
| POST | /designs | Start an AI design job |
| GET | /designs/:id | Status, spec and preview |
| PATCH | /designs/:id | Edit collar, sleeve, hem, fit |
| POST | /designs/:id/refit | Re‑fit to current sizes |
| POST | /designs/:id/pack | Generate the tailor‑pack PDF |
| GET | /packs/:id/pdf | Signed download link |
| DELETE | /me | Erase account and data |
One responsive PWA; install‑to‑home‑screen; offline cache for saved designs.
REST API and workers; pick the language your hires know best.
Row‑level security; OTP through an Indian SMS provider with DLT registration.
Uploads, previews and PDFs behind signed URLs; Cloudflare R2 has no egress fee.
A low‑cost vision model reads attributes; an image model renders the preview. Versions are pinned.
Open‑source parametric patterns (for example FreeSewing; check its licence) or an in‑house engine.
pdf‑lib or PDFKit for 1:1 tiles, test square, registration marks and Devanagari fonts.
On‑device pose tracking with a canvas or three.js overlay; no server cost per session.
About 22 to 26 weeks for the recommended Standard scope with a team of five to seven. Phases overlap; the pilot with tailors is the gate that matters most.
Prove the real problem and settle what “size 32” means before any code.
Responsive screens for phone and desktop, designed once.
Sign up, sizes and profile working on a real staging site.
Turn a messy screenshot into a validated garment spec and a preview.
The accuracy core: sizes in, checked pieces and a 1:1 PDF out.
Camera opens, outfit appears on the user, nothing leaves the device.
Stitch real garments, measure them, fix the patterns.
Make it safe, fast and lawful on every device.
Staged roll‑out, watch cost and fit, keep improving.
Three scopes, priced from role rates and person‑weeks. Each range runs from a freelance‑led team to a studio; a full agency quote is shown beside it.
Prove the loop with real users.
₹24–41 lakhbuild, before 18% GST · agency quote ≈ ₹68 lakh
What we would launch with.
₹27–47 lakhbuild, before 18% GST · agency quote ≈ ₹77 lakh
Scale‑up polish.
₹34–59 lakhbuild, before 18% GST · agency quote ≈ ₹98 lakh
All build figures include a 15% contingency and exclude running costs, GST and your own time.
Tick the modules you want and choose who builds it. The total updates from the same rates and effort used above.
| Role | Freelance | Studio | Agency | Person‑weeks (Standard) |
|---|
Rates are rounded from 2026 India rate cards, which quote about ₹700–1,400 per hour for mid‑level local engineers and $55–90 per hour for mid‑level agency work. A person‑week is 40 hours.
| Active users | 1,000 | 10,000 | 50,000 |
|---|---|---|---|
| Hosting, database, storage, SMS, email, monitoring | ₹5.7k | ₹14.7k | ₹49k |
| AI, lean path (≈ ₹4.4 per design) | ₹13k | ₹1.3 lakh | ₹6.6 lakh |
| AI, balanced path (≈ ₹8.8 per design) | ₹26k | ₹2.6 lakh | ₹13.2 lakh |
| Total, lean path | ₹19k | ₹1.5 lakh | ₹7.1 lakh |
| Total, balanced path | ₹32k | ₹2.8 lakh | ₹13.7 lakh |
Assumes 3 designs per user per month and one 20% contingency on infrastructure. AI is over 90% of the bill, so free designs are capped and extra ones sold as credits. Live try‑on runs on the device and adds almost nothing.
Not a quote. Prices vary by vendor, city and scope, and AI prices change often. Re‑check the model price pages before launch.
Sources: Second Talent India rate card · Project Cost Estimator · eCorpIT AI app cost · Riolabz app cost bands · PerfectionGeeks AI stylist cost · Gemini API pricing · FASHN API pricing · MSG91 OTP pricing · Supabase · Cloudflare R2.
Start with the pilot: a handful of boys, a few tailors, and one sample pack. Everything else follows from what fits.
The geometry is exact: every seam length is computed and checked, and every sheet prints with a 10 cm test square so you can confirm scale. The fit on your body depends on your real measurements, the fabric and the stitching. That is why the plan includes a tailor pilot and a fit‑feedback loop, and why accuracy numbers on this page are targets until they are measured.
A photo shows style, not measurements. The AI reads the style (collar, sleeve, length, fit, pockets) into a structured spec. Your sizes supply the numbers, and a pattern engine draws the pieces. If the photo is unclear, the app asks you instead of guessing.
Brands use different labels: chest in inches, collar, or boys’ sizes. The sign‑up screen explains each one next to the field and offers a quick way to measure. Pant size is the waist in inches. Adding height or a favourite shirt’s measurements makes the result closer.
No. Live try‑on is designed to run on your device. The plan includes a network‑log check in Phase 5 to prove it. Only the outfit you save and your sizes are stored.
The product is meant for boys and young men, so it is designed around India’s data‑protection rules for children: a parent or guardian’s verifiable consent, no behavioural tracking and no targeted ads for under‑18 users. Counsel should review the final flow before launch.
Yes. The pack is written so that any tailor can cut and stitch from it. It replaces the vague “make it like this photo” conversation with measurements, pieces and a stitching order.
It recreates the cut and colours of a garment, not logos or protected prints. Those areas become plain, or you can add your own artwork.
It is one web app that adapts from small phones to wide desktops, and installs to a phone’s home screen. Live try‑on needs a camera and a secure (HTTPS) page. If a camera is not available, you can try on from a photo instead.
The Standard scope is estimated at ₹27–47 lakh to build, with a monthly running cost of roughly ₹32 thousand at 1,000 active users, mostly AI. See the cost section to build your own estimate.
