Input: one garment reference and a body/light direction
Output: model and detail images in one set
Aspect: 3:4 and 1:1
Build log
1. Set the rule
I opened the one clean SKU photo and one selling point, locked the exact product shape, label, color, and material, and wrote down what could not change.
2. Run one minimum version
I made one small model and detail images in one set first. The garment shape drifted, so I locked the garment reference and reduced pose complexity.
3. Build the focused variant
I reused the same reference and changed only one model pose and one product close-up.
4. Review and export
I checked text, edges, shadows, color, and final crop at full size, then exported the final 1:1 and 4:3.
Complete execution record
Task boundary and source gate
Treat AI fashion model product photo as a deliverable, not a definition. The job receives one garment reference and a body/light direction and owes model and detail images in one set in 3:4 and 1:1. Write the rejection reason first, then turn it into the first hard constraint.
Do not start until the source photo has a visible product code, a note on material and color, the approved claim, and a named target module. Check packaging damage, sticker state, lighting direction, and background rights before generation. Do not invent certifications, competitor claims, price windows, or a model using the product in an unsafe way. If the seller cannot prove it, the image does not carry it.
The imgmov route
Open Asset Library first, attach the master SKU as the locked reference, then build the module or shot around that reference. Keep offer copy, platform text, and CTA outside the visual prompt. In imgmov, the approved image becomes the source of truth and generation supplies only the requested scene, angle, or motion.
Keep platform requirements outside the model prompt. The prompt can describe a scene; a named checklist records the crop, claim, consent, and console check that make it deliverable.
Lock the variables that cannot drift
Lock product silhouette, label spelling, logo position, material, primary color, accessory count, contact shadow, and the platform safe area. Scene, props, and motion may change; these fields may not.
Put the locked fields at the top of the prompt, not at the end: The practical move is to lock the exact product shape, label, color, and material and change only one model pose and one product close-up. Save that prompt with the reference, ratio, and model so the next run starts from a decision instead of a guess.
Step-by-step build path
1) Set the rule: I opened the one clean SKU photo and one selling point, locked the exact product shape, label, color, and material, and wrote down what could not change. Leave one checkable artifact from this step; do not start the next until it exists. 2) Run one minimum version: I made one small model and detail images in one set first. The garment shape drifted, so I locked the garment reference and reduced pose complexity. Leave one checkable artifact from this step; do not start the next until it exists. 3) Build the focused variant: I reused the same reference and changed only one model pose and one product close-up. Leave one checkable artifact from this step; do not start the next until it exists. 4) Review and export: I checked text, edges, shadows, color, and final crop at full size, then exported the final 1:1 and 4:3. Leave one checkable artifact from this step; do not start the next until it exists.
Change one named variable per rerun: prompt, reference, camera, duration, model, ratio, or export crop. If two variables change together, a better output cannot be reused because nobody knows which fix worked.
Review gates and evidence
Review at 100%. Read small text, count buttons and parts, compare color against the source, inspect the four corners, and overlay the platform crop. A thumbnail is only a concept check, not a marketplace check.
Keep source, approved wording, rejected version, correction, credit cost, and final crop next to the asset. The record exists so the next reviewer can reproduce the decision without a chat thread.
Failure diagnosis and retry ladder
Name the failure first: label redraw, color drift, shape drift, missing accessory, wrong contact shadow, or cropped CTA. Stop prompting for more letters, reduce the variable count, return to the reference, and only then upgrade the model.
Image retries have a ladder. Start with Agnes Image 2.5 for the cheap proof. Use SDXL Inpainting at 3 credits when only a masked region is wrong. Use Seedream 5.0 Lite or Nano-Banana 2 at 5 credits for a more complex rebuild. GPT Image 2 starts at 2 credits for Standard 1K and rises to 80 by quality and resolution. Name the failure before changing rungs.
Versioning and handoff
Deliver by module and ratio, not one stretched master. Name files with SKU, module, crop, version, and approval state. Keep the rejected render and the correction line in the same folder.
The folder is boring on purpose: source, approved reference, generation settings, caption file, platform cuts, QA screenshots, and one correction line. A reusable asset is boring in the right way.
Why this is not a generic answer
Editing is not detached from production. Object detection, masks, reference images, version history, Canvas dependencies, and the final generation set stay together. A standalone generator can invent a scene; it will not preserve the source asset, the rejected versions, and the reason the approved crop won.
Do not invent certifications, competitor claims, price windows, or a model using the product in an unsafe way. If the seller cannot prove it, the image does not carry it. If the request only asks what ai fashion model product photo means, a search page is faster. This page is useful when someone must deliver model and detail images in one set under real constraints.
Frequently asked questions
What is the one rule I keep repeating?
The practical move is to lock the exact product shape, label, color, and material and change only one model pose and one product close-up.
The workflow is specific about its stopping point: it starts with one garment reference and a body/light direction and stops at model and detail images in one set. In imgmov, upload the source, lock it as a reference, set 3:4 and 1:1, run one proof, and save the passing settings. If a reviewer cannot tell which version is approved, the workflow has failed even when the media looks good.
What do I check before export?
Shape, label, color, edges, and corner pixels.
Image retries have a ladder. Start with Agnes Image 2.5 for the cheap proof. Use SDXL Inpainting at 3 credits when only a masked region is wrong. Use Seedream 5.0 Lite or Nano-Banana 2 at 5 credits for a more complex rebuild. GPT Image 2 starts at 2 credits for Standard 1K and rises to 80 by quality and resolution. Name the failure before changing rungs.
What makes this different from a generic generator?
I start from one garment reference and a body/light direction and keep model and detail images in one set consistent, instead of inventing a new scene each run.
The check is not “does it look AI-nice?” It is: Review at 100%. Read small text, count buttons and parts, compare color against the source, inspect the four corners, and overlay the platform crop. A thumbnail is only a concept check, not a marketplace check. Then the source, approved copy, rejected version, correction, and final crop stay in the same handoff folder.
What should I do if the first render drifts?
Cut the motion, lock the reference, and rerun one small version before batching.
Name the failure first: label redraw, color drift, shape drift, missing accessory, wrong contact shadow, or cropped CTA. Stop prompting for more letters, reduce the variable count, return to the reference, and only then upgrade the model. Change one named variable, keep the old version, and record the credit cost. If the same failure repeats, fix the reference or scope rather than asking the prompt for forgiveness.