AI image generation
Beta
AI image generation produces product images from a text description. It’s an opt-in beta feature; output quality and the underlying model may change as the feature matures.
In this guide:
- Open image generation
- Write a good prompt
- Pick a style
- Save and use the image
- Stay within Amazon’s image policy
Prerequisites
- Beta is opt-in — toggle Settings → AI → Beta features → Image generation to enable.
- AI usage quota.
Step 1: Open image generation
Go to AI → Image generation in the sidebar (only visible if Beta is enabled).
Step 2: Write a good prompt
Good prompts are concrete:
- ✅ “A 2-pack of stainless steel insulated water bottles, white background, studio lighting, top-down view”
- ❌ “A nice product photo”
Mention:
- The subject in detail (color, material, shape, count).
- The background (white, lifestyle scene, gradient).
- The angle (top-down, 3/4, profile).
- The lighting (studio, natural, dramatic).
The longer and more concrete, the better.
Step 3: Pick a style
Style presets above the prompt:
- Listing hero — clean white background, studio lighting, ready for Amazon’s main-image rules.
- Lifestyle — product in a real-world scene.
- Ad creative — bolder composition, room for headline text.
- Social — square format optimized for Instagram / Facebook.
Step 4: Save and use the image
Each generation produces 4 variants. Click any variant to expand. Then:
- Download — saves to your computer.
- Save to library — keeps the image in your Hilal asset library.
- Use in listing — sends to Inventory → Products to attach as a SKU image (subject to Amazon’s rules — see below).
Step 5: Amazon’s image policy
Amazon has strict rules for listing images:
- Main image must be on a pure white background.
- No text, watermarks, or logos.
- Product must fill ≥85% of the frame.
- No props that aren’t sold with the product.
The Listing hero preset attempts to comply, but always review before using as your Amazon main image. AI-generated images are not exempt from Amazon’s policy.
Limitations during Beta
- The model can struggle with text on packaging — review carefully.
- Hands and small repeating details (logos, dials) sometimes look “off.”
- Scenes with multiple distinct objects can confuse the model.
Troubleshooting
- Image looks off. Regenerate with a more concrete prompt, or pick a different variant.
- “Beta not enabled.” Toggle it on at Settings → AI → Beta features.
- Quota exceeded. Image generations use more AI quota than text. Check your usage panel.