dlazy-ai/ecommerce-skills

to-3d — Generate 3D Garment Images from Flat-Lay Photos

Transform a garment photo laid flat on a table into a dimensional 3D image — as if worn by an invisible person (known in the industry as a ghost mannequin / invisible model).

Use case: Flat-lay photos are cheap but look cheap. Real model photos are expensive and unsuitable for all categories. 3D garment images are the middle ground — offering texture, showcasing fit, without model costs or portrait rights issues.

---

Example Output

Input: Flat-Lay Photo
garment-flatlay.jpg — Olive green cable-knit crewneck sweater flat-lay, 800×800

Actual command executed:

dlazy gpt-image-2 \
  --prompt 'Turn this flat-lay garment photo into a dimensional 3D ghost-mannequin product shot. The olive-green cable-knit crewneck sweater must gain realistic volume: filled shoulders and chest, sleeves with natural bend, visible interior of the collar, soft self-shadow under the hem, as if worn by an invisible mannequin. Keep the garment 100% faithful: same olive-green colour, same cable-knit and diamond stitch pattern, same ribbed collar and cuffs, same woven cuff label. Clean seamless light-grey studio background, soft top light, sharp fibre detail. No mannequin, no person, no text.' \
  --images docs/to-3d/garment-flatlay.jpg \
  --size 1024x1024 --quality medium --imageFormat jpeg \
  --save docs/to-3d/example-output.jpg

Output

example-output.jpg — 1024×1024. Shoulders and chest are filled, sleeves naturally bent, collar opening reveals inner ribbing and back lining, hem has a self-weight shadow; cable-knit and diamond stitch pattern, olive-green colour, ribbed structure and right sleeve woven label remain unchanged, no mannequin or body visible in the frame.

---

1. Capability Boundaries

Capability	Description
Dimensional Filling	Shoulder and chest filling, sleeve bending, collar interior visibility, hem self-weight shadow
Garment Types	Tops / Bottoms / Dresses / Outerwear / Underwear / Kids' Wear / Homewear
Shape Control	Reference image (copy a specific 3D shape) or custom prompt
Texture Enhancement	On/Off switch; when enabled, surface details of knits, fleece, leather become clearer
Generation Aspect Ratio	1:1 (square main image) / 3:4 (vertical detail image)

Does NOT do: Change colour, weave, print, or ribbing structure; generate human bodies or faces; alter the actual garment fit proportions.

---

2. Input Material Requirements

Check these hard constraints before generating:

• Size: 20KB ~ 15MB
• Resolution: Greater than 400×400
• Format: jpg / jpeg / png / webp

Recommended Inputs (✅): Single flat-lay garment on a plain background, completely flat, no wrinkles or folds, full front view, with clear collar and hem boundaries.

Inputs That Clearly Degrade Results (❌)

Issue	Description
Folded Placement	Sleeves folded on the body; the model cannot calculate sleeve length
Heavy Wrinkles	Wrinkles will be interpreted as structure and deform the shape
Already on a Body	Use one-shot for images already on a person
Set Items in Same Frame	Separate into single items and process individually

---

3. Three Control Points for the 3D Shape

Control Point	Write in Prompt	What Happens If Not Written
Fill Level	filled shoulders and chest with realistic volume, natural fabric drape	Fills up like a balloon, proportions distorted
Sleeve Pose	sleeves with a natural bend, hanging slightly forward and away from the body	Sleeves stick straight to the sides
Collar Interior	visible interior of the collar showing the inner facing	Collar is a flat hole, the most obvious giveaway

Add two environmental constraints:

soft self-shadow under the hem and inside the sleeves
No mannequin, no person, no visible support — the garment must appear worn by an invisible body.

The phrase No mannequin, no person must be included, otherwise the model often generates a mannequin torso or half a body.

---

4. Tool Invocation

This skill uses dLazy's gpt-image-2 (image editing model; inferring volume from 2D requires strong geometric understanding while strictly preserving weave and colour).

Invocation Methods

Two equivalent methods, choose one. The unified entry point automatically selects the backend, retries on failure, creates directories, saves to disk, and estimates cost:

# A. Unified entry point (recommended): switchable backend, add --dry-run for cost-free test-run
node scripts/gen.mjs --task to-3d \
  --prompt '<See Prompt Template below>' \
  --images <in order from table below> \
  --save output/to-3d-<sku>.jpg

# B. Directly use dLazy CLI (when you don't want Node dependencies, equivalent effect)
dlazy gpt-image-2 --prompt '...' --images ... --save output/to-3d.jpg

Parameter Conventions (fixed usage for this skill)

Parameter	Value	Reason
--images	[flat-layer image]; with shape reference [flat-layer, shape-reference]	Order corresponds to image 1 / 2 in prompt
--size	1024x1024 (square main image) / 1024x1536 (vertical longs, dresses)	Corresponds to platform's 1:1 / 3:4 ratios
--quality	high (equivalent to "texture enhancement on") / medium (off)	Knits, fleece must use high
--imageFormat	jpeg	Universal e-commerce listing format
--batch	2 ~ 3	Volume inference has randomness
--save	docs/to-3d/output-<sku>.jpg	Directly saves to disk

Command Examples

# basic call
dlazy gpt-image-2 \
  --prompt 'Turn this flat-lay garment photo into a dimensional 3D ghost-mannequin product shot. Filled shoulders and chest with realistic volume, sleeves with a natural bend, visible interior of the collar, soft self-shadow under the hem. Keep the garment 100% faithful: same colour, same stitch pattern, same ribbing, same label. Clean seamless light-grey studio background. No mannequin, no person, no text.' \
  --images docs/to-3d/garment-flatlay.jpg \
  --size 1024x1024 --quality high

# complex call: copy a specific 3D shape + vertical + generate 3 images for selection
dlazy gpt-image-2 \
  --prompt 'Turn image 1 (a flat-lay garment) into a dimensional 3D ghost-mannequin product shot, copying the volume, posture and camera angle of image 2. Filled shoulders and chest with realistic volume and natural fabric drape, sleeves with a natural bend hanging slightly forward, visible interior of the collar showing the inner facing, soft self-shadow under the hem and inside the sleeves. Keep the garment 100% faithful to image 1: same colour, same knit structure, same print placement, same ribbing and label. Clean seamless light-grey studio background, soft top light, sharp fibre detail. No mannequin, no person, no visible support, no text.' \
  --images docs/to-3d/garment-flatlay.jpg docs/to-3d/volume-ref.jpg \
  --size 1024x1536 --quality high --imageFormat jpeg \
  --batch 3 --save docs/to-3d/output-sku001.jpg

# estimate cost without actually running
dlazy gpt-image-2 --dry-run --prompt '...' --images a.jpg --size 1024x1024 --quality high

Further Reading

What to Look Up	Where
Authentication, multi-backend configuration, output structure, error codes	references/provider-cli.md
All available parameters for gpt-image-2	references/model-flags.md
All options for the unified entry point	node scripts/gen.mjs --help

5. Prompt Template

Turn this flat-lay garment photo into a dimensional 3D ghost-mannequin product shot.

The [Category + Color + Fabric] must gain realistic volume:
filled shoulders and chest, sleeves with a natural bend,
visible interior of the collar showing the inner facing,
soft self-shadow under the hem and inside the sleeves,
as if worn by an invisible body.

Keep the garment 100% faithful: same [Color], same [Weave/Texture], same [Print/Pattern Placement],
same [Ribbing/Collar/Cuff/Hem Structure], same [Tag/Label].

Clean seamless [Background Color] studio background, soft top light, sharp fibre detail.
No mannequin, no person, no visible support, no text.

Correction Sentences for Specific Issues

Issue	Append to End of Prompt
Mannequin or half-body appears	Absolutely no mannequin, torso, neck, arms or hands anywhere in the frame — only the garment.
Overfilled, proportion widened	Keep the original garment proportions: body width, sleeve length and hem width must match image 1 exactly. Add volume, not size.
Collar is a flat hole	The collar opening must show the inner facing and the inside back of the garment, with correct depth and shadow.
Weave smoothed out	Change --quality high, append resolve individual knit loops and yarn twist
Shadow direction inconsistent	Single soft top-left light source; one consistent contact shadow beneath the garment.

---

6. Execution Flow

1. Validate Input: Completely flat, no folds, no heavy wrinkles, single item, clear boundaries.
2. Write Three Control Points (Fill / Sleeve Pose / Collar Interior) + Two Environmental Constraints (Self-Weight Shadow / No Mannequin).
3. Write Fidelity Items: Colour, weave, print placement, ribbing structure, tag.
4. Select Aspect Ratio and Quality: Knits, fleece always use --quality high.
5. --dry-run Estimate → Run → --batch 2~3 Select Image, Save to docs/to-3d/.
6. Quality Check: Check for human body generation, proportion widening, correct collar interior, preservation of weave.

---

7. Frequently Asked Questions

Observation	Cause	Solution
Mannequin/half-body appears	Missing prohibition sentence	Append strengthened prohibition sentence (Section 5)
Proportion widened	Only mentioned adding volume, not preserving proportions	Append proportion preservation sentence: Add volume, not size.
Collar is a flat black hole	Interior not described	Append collar interior sentence
Sleeves stick straight to sides	Sleeve pose not described	Append hanging slightly forward and away from the body
Weave blurry	--quality medium	Switch to high; or chain with material-enhancement
Poor result with folded image	Input violation	Retake as a completely flat flat-lay photo

---

Tips

Visit https://dlazy.com for more information.