01 · Baseline Failure Mode
Where Off-the-Shelf Workflows Broke
Supplier smartphone cutouts pasted onto stock backgrounds converted poorly, while SaaS AI background tools warped nutrition facts and shifted brand colors.
Multi-Brand Amazon FBA Aggregator · 7 min technical case study
Scaling 84 Amazon A+ Listing Visuals with n8n & Locked ComfyUI Relighting
How Insiyak replaced a 6-week physical tabletop shoot backlog with an automated n8n + ComfyUI relighting factory preserving 100% of product label pixels.
| Benchmark KPI | Before Insiyak | After Deterministic Pipeline |
|---|---|---|
| Per-SKU Cost | $420 studio shoot | $38 automated + QC |
| 80-SKU Turnaround | 6 weeks | 4.5 business days |
| Label Typography Drift | 34% defect rate in SaaS AI | 0.0% (pixel-locked) |
02 · Production Systems Architecture
How Insiyak Solved It
We engineered an n8n-triggered headless ComfyUI DAG that extracts a sub-pixel product mask, generates a physically grounded countertop environment via Depth ControlNet, computes directional light-wrap with IC-Light, and composites the untouched original SKU pixels back on top.
Pipeline DAG Node Breakdown
Stage-by-stage execution topology.
Stage 01
n8n Webhook & Google Drive SKU Watcher -> JSON Job Builder
Stage 02
RMBG-2.0 / SAM2 Sub-Pixel Alpha Matte & Edge De-Fringe
Stage 03
Flux.1 Dev + Depth ControlNet Countertop Scene Synthesis
Stage 04
IC-Light Ambient & Rim Relight Harmonizer + Contact Shadow Pass
Stage 05
Non-Destructive Pixel-Locked Label Composite + 4K sRGB Export
Principal Architect Takeaway
Never let diffusion denoise real packaging typography—use generative AI for environmental light transport and shadow grounding around a locked foreground mask.
Deploy This Architecture for Your Team