name	product-photo-studio
description	Transform product photos via Picsart gen-ai — six modes.
license	MIT
metadata	{"author":"Picsart","version":"1.0.0","hermes":{"category":"creative","tags":["picsart","product-photos","ecommerce","catalog","mockups","variants"]}}

Product Photo Studio

A single skill that covers every "transform a product photo with AI" workflow Picsart's gen-ai CLI supports. Use this whenever the user has product photography (single packshot, full catalog, or one artwork) and needs it re-rendered, re-staged, or fanned out into variants. Replaces six narrower skills with one entry point and six mode references.

Input: one or more product photos. Output: styled, composed, or fanned-out variants ready for PDPs, marketplaces, ads, mockup listings, or campaigns.

When to Use

Pick the mode that matches the task. If the user's request maps to any of these, this skill is the right one:

Mode	Trigger phrases	Reference
bulk-restyle	"catalog styling", "consistent background across SKUs", "PDP-ready staging at scale"	`references/modes/bulk-restyle.md`
compose	"lifestyle compose", "product in context", "put this product in a scene"	`references/modes/compose.md`
seasonal	"seasonal refresh", "holiday catalog", "Christmas / summer / Black Friday catalog"	`references/modes/seasonal.md`
variants	"color variants", "material variants", "colorway fan-out", "size chart re-render"	`references/modes/variants.md`
reshoot	"batch reshoot", "regenerate catalog", "enterprise catalog re-render with brand rules"	`references/modes/reshoot.md`
mockups	"product mockup", "POD mockup", "Etsy / Shopify listing mockup", "in-hand render"	`references/modes/mockups.md`

If the user's task involves video, characters, or persona generation, this is the wrong skill — see gen-ai-use, gen-ai-persona-creation, gen-ai-use.

Prerequisites

Picsart gen-ai CLI installed and authenticated:

# Install (signed binary, recommended)
curl -fsSL https://picsart.com/gen-ai-cli/install.sh | bash

# Authenticate
gen-ai login
gen-ai whoami    # verify

Per-mode prerequisites (image counts, brand files, manifests) are documented inside each mode reference. Always confirm pricing before a bulk run:

gen-ai pricing --model <model> --count <N>

How to Run

Identify the mode from the user's request using the table in When to Use.
Load the corresponding mode reference: Read references/modes/<mode>.md.
Follow the procedure described there — interview, manifest, generate.
Return to this SKILL.md only when switching modes mid-task.

Quick Reference

# Single image (compose / mockups)
gen-ai generate --model <model> --image input.jpg --prompt "<prompt>"

# Batch (bulk-restyle / seasonal / variants / reshoot)
gen-ai batch --manifest manifest.json

# Estimate cost before running
gen-ai pricing --model <model> --count <N>

# Browse available models
gen-ai models

Manifest patterns, model recommendations, and per-mode best practices live in the individual mode references.

Procedure

Always follow the same outer loop regardless of mode:

Interview — confirm: which mode, how many inputs, target output format(s), brand constraints, deadline.
Manifest — assemble a JSON manifest (single-shot inline, or batch file). Each mode reference has a template.
Estimate — run gen-ai pricing before committing. Surface the total to the user.
Generate — invoke gen-ai generate or gen-ai batch. Stream progress.
Verify — open the output directory and confirm the expected files exist with the expected dimensions.
Hand off — drop into the configured Drive folder, marketplace feed, or deliverable zip.

Pitfalls

Mode-specific pitfalls (e.g. "shadow direction drifts across variants", "seasonal overlays expire", "brand.md gate blocks SKU-XYZ") live inside each mode reference. Shared pitfalls:

Never skip the pricing estimate. Bulk runs across thousands of SKUs can rack up real cost.
Always namespace outputs by client/run — don't write into a global output dir, you'll lose track of which batch produced which assets.
Respect brand governance. If the user has an enterprise-brand-governor-style brand.md in scope, the manifest must reference it.
Don't switch models mid-batch. Pin the exact model version per run for consistency; see enterprise-pinned-registry.

Verification

After any run:

# Confirm output count matches expected
ls -1 outputs/<run>/ | wc -l

# Spot-check one output's dimensions
gen-ai inspect outputs/<run>/<sample>.jpg

If anything looks off, re-run with --debug and consult the mode reference's "Common pitfalls" section.