AI-Powered Interior Visualization

Visualize any room with AI texture preview in seconds

Production-grade SaaS for wallpaper & PVC panel dealers. Auto wall detection → texture mapping → photorealistic output.

Live Demo Application Response
Interactive Demo
Try the AI Preview Engine

Pick a room, select a wall texture, and simulate the full SAM + ControlNet + SDXL pipeline. Textures applied via Canvas Pattern API — deterministic, tile-accurate.

1
2
3

Room Scene

🛋️Living Room
📺TV Lounge
📚Study / Library
🪑Dining Room
🛏️Bedroom

Wall Texture

Bamboo
Marble
Oak Wood
Linen
Carbon
Concrete

Select a room & texture,
then hit Generate

Technical Blueprint
Proposed Architecture
 Frontend — React SPA (Vite)
Dealer Portal
→
Image Upload
→
Before/After Preview
→
WhatsApp Share
 Backend — FastAPI (AWS EC2)
FastAPI + JWT
→
Multi-tenant Auth
→
PostgreSQL (RDS)
Celery Worker
→
Redis Broker
→
AWS S3
→
CloudFront CDN
 AI Inference — RunPod GPU (Serverless)
SAM vit_h — Wall Segmentation
→
ControlNet Canny+Depth
→
SDXL Turbo
Perspective Warp
→
Poisson Blend
→
Watermark Inject
→
PNG/WebP Output
 GPU Specs
A100 40GB — recommended RTX 4090 24GB — minimum SDXL: 10–14GB VRAM SAM vit_h: 6–8GB VRAM ControlNet: +3GB ~18–28s end-to-end RunPod serverless = zero idle cost

SAM — Wall Detection

Meta's Segment Anything (vit_h). Fully automatic mask generation, no manual input. Point-prompt fallback for complex geometry. ~3–5s on A100.

ControlNet + SDXL Turbo

Canny + Depth conditioning preserves room structure. SDXL Turbo for 4-step inference speed. Interior LoRA fine-tune for photorealism.

Multi-Tenant SaaS

Per-dealer PostgreSQL schemas, Stripe subscription control, usage counters, admin CRUD dashboard, watermark injection for trial accounts.

AWS + RunPod Infra

FastAPI on EC2, RDS Postgres, S3 + CloudFront CDN. RunPod serverless GPU scales to zero when idle. All infra under client accounts.

Delivery Plan
8-Week MVP Timeline
Week 1–2 · Milestone 1
Wall Detection Working Demo
SAM vit_h pipeline live on RunPod. Upload room image → segmented wall mask overlay. Client reviews on 10+ real samples. Deliverable: video + live API endpoint.
Week 3–4 · Milestone 2
Realistic Texture Application Demo
ControlNet + SDXL connected to SAM masks. Perspective-aware placement, Poisson blending. Deliverable: before/after set across 5 room types.
Week 5–6 · Milestone 3
Full SaaS Web Structure
React + FastAPI live. Multi-tenant auth, product catalog, preview history, admin dashboard, subscription + watermark logic.
Week 7 · Milestone 4
Full Integration + Testing
End-to-end QA: load testing, edge-case wall shapes, concurrency, WhatsApp share, download/watermark. Target: <40s per generation.
Week 8 · Milestone 5
Deployment + Documentation
Production deploy under client AWS + RunPod. Source code, arch diagram, DB schema, env vars, AI model guide, deployment runbook. 3-month bug support begins.

Application Response

Full technical answers to every requirement in the job posting.

 Models I'll Use

SAM vit_h — automatic wall segmentation, zero manual masking.
ControlNet Canny + Depth — preserves room geometry during inference.
SDXL Turbo — 4-step photorealistic texture synthesis at speed.
Optional: custom interior LoRA fine-tuned on client's product catalog for brand-accurate output.

 GPU Requirement

Minimum: NVIDIA RTX 4090 24GB — handles full SDXL + SAM + ControlNet stack concurrently.
Recommended: NVIDIA A100 40GB via RunPod serverless — faster batching, concurrent user headroom.
Cost per generation ≈ $0.005–$0.02. Zero idle cost with serverless.

 Expected Generation Time

SAM wall detection: ~3–5s  |  ControlNet + SDXL (4-step): ~12–16s  |  Perspective + Poisson blend: ~2–3s  |  S3 I/O: ~1–2s
Total: ~18–26s — well within the 40s requirement.

 Architecture Overview

React SPA → FastAPI (EC2) with JWT multi-tenant auth → Celery + Redis task queue → RunPod serverless GPU (SAM → ControlNet → SDXL → Poisson blend) → S3 → CloudFront → dealer browser.

PostgreSQL on RDS: dealer accounts, product catalogs, preview history, usage counters, subscription state. Admin panel in FastAPI + React. All infra provisioned under client's AWS + RunPod accounts. GitHub private repo, scoped developer access.

This page is the portfolio demo — full-stack AI SaaS capability. Source available on request.

Submit Application with this Portfolio Link