Jul 21, 2026· 5 min read
Phone Photos + AI: The Two-Step Studio for Salla & Shopify Merchants
How to shoot garment photos on your phone and let AI generate on-model studio images: the capture routine, the mistakes to avoid, and the 30-second step two.

Can phone photos really replace a product photo studio?
Yes — for exactly half of the job, and that half is the one you can do tonight. A recent phone in good daylight captures the truth of a garment: its real color, texture, stitching, and construction. AI generation handles the other half — the model, the pose, the scene, the lighting, and every export ratio — in about 30 seconds per image. That split is the two-step studio: truthful capture on your phone, studio-grade output from generation, with nothing to buy in between.
Why the two-step split works
Phone cameras have become excellent documentary instruments. In indirect daylight, a phone records fabric weave, dye color, and seam detail accurately enough to serve as a reference. What a phone cannot do is everything a studio charges for: hire a model, build a set, light a scene, and keep fifty products visually consistent across shooting days.
Generation tools are the exact mirror image. They produce models, scenes, and studio lighting on demand — but they need a source of truth about your specific garment, or they invent one. Rokon joins the two halves: your phone photos become a multi-image reference that preserves the actual stitching, logo, print, and silhouette, while the system generates only the model, pose, and background. The phone supplies the truth; the AI supplies the production.
This matters most for stores that sell what they photograph. A supplier's render or a stock image shows a garment; your phone photo shows your garment — the one the customer will receive. Generation built on that photo keeps the promise.
Step 1 — The 10-minute capture routine
You need a window, a plain surface, and a phone. In the Gulf, avoid harsh midday sun coming straight through the glass — morning or late-afternoon light is softer and truer, especially on black fabric.
- Light: indirect daylight from a window, no ceiling lights on. Mixed light sources shift color.
- Background: a plain floor, wall, or bedsheet in a color that contrasts with the garment. A black abaya on a dark carpet loses its edges.
- Garment prep: steam or iron first. Wrinkles read as design features to a generator.
- Position: flat lay or a simple hanger shot, the full garment inside the frame — no cropped hems or sleeves.
- The set: front, back, and 1–3 detail close-ups — logo, embroidery, buttons, cuff, fabric texture. 3–5 photos per garment is the working standard.
- Phone settings: turn off filters, beauty modes, and aggressive HDR. You want the boring, accurate photo.
Ten minutes per garment is a realistic pace once the corner by the window is set up. A rack of 20 pieces is an afternoon, not a project.
Capture mistakes that ruin generations
| Mistake | What it does downstream | The fix |
|---|---|---|
| Mixed lighting (lamp + window) | Color casts the AI faithfully reproduces | One light source: the window |
| Busy background | Edge confusion, stray artifacts | Plain contrasting surface |
| Wrinkled garment | Creases rendered as permanent design lines | Steam before shooting |
| Cropped hem or sleeve | The generator guesses the missing part | Whole garment in frame |
| Filters or beauty modes | False color the AI cannot un-shift | Shoot clean, no processing |
| Dark piece on dark surface | Lost silhouette, soft edges | Light background for dark garments |
Every one of these errors is invisible at thumbnail size and obvious after generation. The AI amplifies whatever the source photo says — which is why the boring, honest photo wins.
Step 2 — Generation does the studio work
Upload the set, and the studio half begins. Configure the model: gender, regional look — including a Gulf-Arab preset — skin tone, and age, with modest-fashion styling (abaya, thobe, hijab) and face-visibility controls built in. Pick poses from the 20+ pose library, up to 8 per garment on the Business plan.
Generation returns each image in roughly 30 seconds (30–120 seconds at higher resolutions). From one capture set you can produce the full listing gallery: hero shot, alternate poses, and every ratio your channels need — export up to 4K in JPG, PNG, or WebP. And because the model is a saved configuration rather than a booking, the same face and styling can appear on every product you shoot this way — the consistency a phone alone can never deliver across shooting days. Salla merchants publish through the native embedded app installed from the Salla App Store; Shopify merchants do the same through the embedded Shopify app.
What the two-step studio replaces
| Route | Typical cost | Turnaround |
|---|---|---|
| Freelance photographer | $300–$500 per product (SAR ~1,100–1,900) | 1–2 weeks |
| Studio shoot | $1,000–$10,000 (SAR ~4,000–40,000) | 2–3 weeks |
| Phone + Rokon | your phone + ≈ $0.50/image on Pro (≈ SAR 2) | same day |
The turnaround difference compounds around the retail calendar. A Ramadan or White Friday refresh booked with a photographer needs weeks of lead time; the two-step studio turns received stock into published on-model imagery the same week it arrives.
The honest limits
The capture step is a ceiling, not a formality: generation cannot recover detail a source photo never contained, so a blurry, color-shifted capture produces a confident-looking image of slightly the wrong garment. Keep the pre-publish habit of zooming into every output — neckline, buttons, print alignment, logo, hands, colorway — against the physical piece. Heavily embellished fabrics such as sequins and dense embroidery deserve the strictest check, and hero editorial campaigns still favor a physical shoot with a real team. Rokon covers the failure case with a credit refund on faulty output, typically within ~24 hours.
Start with one garment tonight
Pick one piece, spend ten minutes at the window, and run the set through a generation. The free tier includes 5 free generations (150 credits) with no credit card, and you own 100% commercial rights to whatever you publish. If the output survives your own zoom test, the two-step studio is already working.
Frequently Asked Questions
Which phone do I need for the capture step?
Any recent phone works. Light matters far more than megapixels: indirect daylight near a window beats an expensive camera under mixed indoor lighting. Turn off filters and beauty modes so the photo stays truthful.
How many photos should I take per garment?
3–5: front, back, and 1–3 detail close-ups of the logo, embroidery, buttons, or fabric texture. More truthful angles give the multi-image reference more evidence, which raises garment fidelity in the output.
Should I edit my photos before uploading them?
No. Cropping is fine, but skip filters, color grading, and beauty modes — the generator faithfully reproduces whatever color shift you bake in. The accurate, unedited photo is the best source.
How long does it take to get finished images?
About 30 seconds per image, with a 30–120 second range at higher resolutions. A full listing gallery from one capture set — hero, alternate poses, and channel ratios — is realistic within the same hour.
What if a generated image gets a detail wrong?
Zoom-check every output against the physical garment before publishing — neckline, buttons, print, logo, hands, and colorway. Faulty outputs qualify for a credit refund, typically within about 24 hours.
About the author
Rokon Editorial TeamThe Rokon team builds AI fashion-photography tools for Gulf & Saudi e-commerce brands.