virtual-tryon

name	virtual-tryon
description	Virtual try-on workflow. Composites a garment onto a subject using reference images, then optionally animates the result. Trigger when the user wants to see clothing on a person or mannequin.
metadata	{"type":"pattern"}

Trigger condition

Use this pattern when the user provides (or asks to generate) a subject image and a garment image, and wants to see the garment worn by the subject — as a still or animated.

Quality checks on input references

Before planning nodes, verify:

1. Subject image: full-body or torso visible, neutral pose preferred, clear lighting. If no subject is available, generate a character sheet first using t2i.
2. Garment image: flat-lay or on-mannequin, full garment visible, minimal background clutter.

If either reference is absent or poor quality, add a t2i node to generate a suitable reference first. For recurring subjects, generate a character sheet (operation t2i, label "Character Sheet — [name]") to anchor visual consistency across multiple try-on nodes.

Workflow steps

[Step 1 — optional] t2i: generate subject reference (if no canvas image available)
  operation: t2i
  promptIntent: "Full-body neutral-pose subject on clean background"

[Step 2 — optional] t2i: generate garment flat-lay (if no canvas image available)
  operation: t2i
  promptIntent: "Flat-lay product shot of [garment description] on white background"

[Step 3 — required] i2i: composite try-on
  operation: i2i
  promptIntent: "Subject wearing [garment], photorealistic composite, same lighting as subject reference"
  edges: subject_ref from subject node, style_ref from garment node

[Step 4 — optional] i2v: animate the composited result
  operation: i2v
  promptIntent: "Subtle breathing and fabric movement, slow 360 rotation"
  edges: depends_on from step 3

Dependencies

• Step 3 (i2i) depends on both the subject and garment images — wire both as edges (subject_ref, style_ref).
• Step 4 (i2v) depends on step 3 — wire as depends_on.

Common failures

• Low-resolution or heavily cropped subject images produce poor composites. Always use at least a torso-visible crop.
• Busy garment patterns (stripes, checks) can bleed into the subject in i2i. Increase prompt specificity: "sharp plaid pattern, precise seams".
• Skipping the subject reference generation step when the user has no canvas image leads to a generic composite. Generate the subject first.