AI Studio's Create from Scratch flow gives you full control over every aspect of your image — from the product input and model choice to pose, background, and styling. This article walks through each step in detail so you can consistently produce high-quality, on-brand results.
<!-- 📸 HERO IMAGE: The full "Create from Scratch" flow showing all 5 step indicators at the top (Product → Classify → Model → Style & Mood → Review) -->
Before You Start: Getting Your Product Input Right
The quality of your output is directly tied to the quality of your product input. Here's what works best:
Ideal product images:
Clean, well-lit shot against a plain white or neutral background
Full product visible with no cropping at edges
JPG or PNG, minimum 800×800px recommended
Avoid heavily styled or already-on-model images — AI Studio works best with flat lay or ghost mannequin shots
Product URL: If you paste a product URL (from Shopify or another store), AI Studio will extract all product images automatically. Use this when you have multiple angles already hosted online — it saves upload time and gives the model more to work with.
Selecting from My Styles: If you've previously added this product to your style library, select it from My Styles. This preserves classification and product metadata, ensuring consistency across multiple shoots.
Tip: Using 2–4 product images (front + back + detail) gives AI Studio more context and typically produces more realistic garment fit and texture rendering.
<!-- 📸 SCREENSHOT: "Start with your product" screen with the URL input field populated, the "Upload from Device" card, and "Select from My Styles" card — all three options visible -->
Choosing Your Aspect Ratio
Select the output ratio before uploading, as it affects how poses and backgrounds are framed.
Ratio | Pixels (approx.) | Best for |
9:16 (Reels) | 1080 × 1920 | Instagram Reels, TikTok |
1:1 (Square) | 1080 × 1080 | Instagram feed, marketplace |
3:4 (Portrait) | 1080 × 1440 | Product detail pages |
4:3 (Landscape) | 1440 × 1080 | Website banners |
16:9 (Widescreen) | 1920 × 1080 | YouTube, desktop heroes |
Tip: For most eCommerce use cases, 3:4 (Portrait) is the safest default — it works on PDPs, app thumbnails, and most catalogue layouts.
<!-- 📸 SCREENSHOT: Aspect ratio selector showing all 5 ratio options with the 3:4 Portrait option highlighted/selected -->
Step 1: Classify Your Images
After uploading, AI Studio auto-detects your product category and classifies images as Front or Back views.
<!-- 📸 SCREENSHOT: Classification screen showing a dress classified under "Women > Western Wear > Dresses" with Front/Back image cards and the AUTO badge on each -->
When to adjust classification:
If AI Studio misidentifies a side view as Front or Back, drag it to the correct column
If you have a detail shot (e.g., close-up of embroidery), you can leave it in Front — it adds texture context
For single-image uploads, leave the default as-is
Classification is optional but recommended. A correctly classified set improves pose accuracy, especially for asymmetric garments like wrap dresses or one-shoulder tops.
Step 2: Choose Your Model
AI Studio offers 95+ models across different ethnicities, age ranges, and body types.
<!-- 📸 SCREENSHOT: Model selection grid showing a range of model cards — diverse skin tones, age labels (e.g., "25–30"), and the "View all 95 models" button at the bottom -->
Choosing strategically:
Match model demographics to your target audience for higher conversion
For high-fashion editorial shots, look for models tagged in the 20–28 age range with a taller frame
For everyday casual or plus-size lines, filter or browse for representative body types
Use your models — if you've previously created a custom model or have a brand face, select it from Use your models for continuity across campaigns
Tip: You can only select one model per creation run. If you want the same product on multiple models, run separate generations. Each run costs credits independently.
<!-- 📸 SCREENSHOT: "Use your models" view showing custom model cards alongside the standard model library toggle -->
Step 3: Choose Style & Mood
This step has the biggest impact on the final look and feel of your images.
<!-- 📸 SCREENSHOT: Full Style & Mood screen showing Shot Type toggle (Catalog vs Editorial), the pose preset grid, and the Auto-Style toggle -->
Shot Type
Catalog — Studio-lit, neutral backgrounds, product-forward. Best for marketplaces (Amazon, Myntra, Meesho) and PDPs where the product needs to be the clear focus.
Editorial — Lifestyle scenes, mood-driven lighting, contextual settings. Best for brand campaigns, social content, and lookbooks.
Pose Selection
Each pose preset pack contains 4–9 individual poses. Click a preset to preview the poses inside.
Standing full-body — most versatile; works for all garment types
Seated / casual — great for loungewear, denim, and lifestyle shots
Detail/close-up — for tops, accessories, or garments with texture details worth highlighting
Use your poses — if you've saved custom poses or reference poses, access them here
Tip: For a new product launch, run one batch with a standing full-body pose (catalog) and one with an editorial lifestyle pose. You'll get content for both your PDP and Instagram in a single session.
Auto-Style: On vs Off
Auto-Style ON (default): AI Studio automatically selects the background, accessories, and overall aesthetic to match your product. This is the fastest path to good results — use it when you're iterating quickly or don't have specific brand requirements.
Auto-Style OFF: Unlocks three additional controls:
Background — choose from preset environments (studio white, outdoor, café, urban street, etc.) or describe your own
Style Accessories — add props or styling notes (e.g., "minimal jewelry", "layered necklaces", "tote bag")
Additional Info — free-text field for anything specific: "model looking over shoulder", "moody evening lighting", "warm terracotta tones"
<!-- 📸 SCREENSHOT: Auto-Style OFF state showing the expanded Background selector, Style Accessories field, and Additional Info text input -->
Tip: When turning off Auto-Style, be specific but not overly prescriptive. "Warm outdoor café, golden hour lighting" works better than a 5-line paragraph describing every detail.
Step 4: Review & Generate
Before spending credits, confirm your full configuration on the Review screen.
<!-- 📸 SCREENSHOT: Review screen showing the three summary cards (Product with image thumbnail, Model name, Style & Mood preset name) next to the AI Model and Resolution pickers -->
Choosing an AI Model
Model | Best for | Notes |
Auto | First-time use, mixed product types | AI Studio picks the best model for you |
Nano Banana 2 | Photoshoots, images with graphic text or prints | Strong at preserving print patterns |
Nano Banana Pro | High-end fashion, elevated realism | Enhanced detail rendering |
Seedream 4.5 | Balanced quality, strong product fidelity | Good all-rounder for most eCommerce |
Seedream 5.0 Lite | Best-in-class realism, pose accuracy | Supports Ultra (4K); use for hero images |
Model selection rule of thumb: Start with Auto or Seedream 4.5 for everyday use. Move to Seedream 5.0 Lite for final hero images or when you need maximum realism. Use Nano Banana 2 if your product has bold graphic prints or text.
Choosing Resolution
Resolution | Use case | Credits per image |
0.5K | Quick preview / internal review | Lower |
Standard | Social media, app thumbnails | 10 credits |
High | PDPs, campaign assets | 10 credits |
Ultra (4K) | Print, large-format display | 10 credits |
Credits math: Each Full HD image costs 10 credits. A typical session generating 6 poses at Standard resolution = 60 credits. Growth plan users have 300 credits/month. Free plan users have 5 credits/month.
<!-- 📸 SCREENSHOT: Resolution selector showing 0.5K, Standard, High, and Ultra options with the credit cost displayed -->
Click Generate Images to start. Generation typically takes 30–90 seconds depending on resolution and server load. Your images will appear in All Creations when ready.
After Generation: What to Do Next
Once your images are ready:
Download — save directly from All Creations
Edit — use Text Edit to refine details via a prompt, or Visual Edit for targeted area changes
Regenerate — re-run with the same settings to get a different variation
<!-- 📸 SCREENSHOT: All Creations view for a completed job showing the generated image grid with Download, Edit, and Regenerate action buttons -->
Troubleshooting
Garment looks distorted or doesn't fit the model properly: Use a cleaner product image with less background clutter. Switching from Auto to Seedream 5.0 Lite often improves garment fit accuracy.
Prints or patterns aren't rendering correctly: Switch to Nano Banana 2 or Nano Banana Pro — these models are specifically tuned for graphic prints and pattern preservation.
Background bleeds into the garment: This usually happens with very light-coloured products on white backgrounds. Try adding a slight shadow or border to your product image before uploading.
Pose looks unnatural: Try a different pose preset from the same pack, or toggle Auto-Style OFF and add a description in the Additional Info field to guide the pose direction.