What is AI Image Generation?
AI Image Generation lets you transform your existing product images using AI-powered prompts. Write a text prompt describing what you want (e.g., "place the product on a marble countertop with soft natural lighting"), and Marpipe will generate a new version of every targeted product image in your catalog.
Generated images are organized into AI image fields — think of them like custom columns that hold your AI-generated creative. You can create multiple fields, each with its own prompts and targeting.
Getting Started
1. Create a new AI image field
Navigate to Generative AI in the main navigation. Click New AI image field, give it a name, and click continue.
Tip: Field names can't match any existing internal field name. Spaces are automatically converted to underscores.
2. Write your prompt
You'll land in the playground — this is where you craft and preview your prompts. On the left panel, you'll see:
AI Agent — Select the AI model to use for generation
Product Image — Choose which feed column contains your source images (usually
image_link)Prompt — Write your text prompt describing the image transformation you want
Target SKUs — See how many products in your catalog will be affected
3. Preview your results
Click Generate Previews to see how your prompt looks across a sample of products. Use Shuffle All to preview different products from your catalog.
Take your time here — tweak the prompt, regenerate previews, and shuffle through products until you're happy with the results.
4. Save and generate
When you're satisfied with your prompt, click Save and continue. You can then generate images for your entire catalog from the AI image field's details page.
The details page shows you all prompts in the field, their statuses, and SKU counts. Click Generate to kick off a full render — you'll see the total image count and cost before confirming.
Note: You can leave the page while images are generating. You'll be notified when generation is complete.
Using Prompt Presets
Don't want to write prompts from scratch? Browse the Presets Library for curated, ready-to-use prompts. Presets cover a wide range of styles — lifestyle shots, editorial looks, studio setups, artistic effects, and more.
You can filter presets by tag to find what fits your products. When you select a preset, it pre-fills your playground so you can preview it immediately and customize from there.
Working with Multiple Prompts
Each AI image field can contain multiple prompts. Prompts are applied in a hierarchy (similar to how rules work in custom fields) — this lets you use different creative treatments for different product segments within the same field.
From the AI image field details page, you can:
Add new prompts
Reorder prompts to set priority
See SKU counts and overlap between prompts
Rename or delete individual prompts
Reviewing Generated Images
After generation completes, open the View Images page to review your results. Here you can:
Browse all generated images (50 per page)
Filter by individual prompt using the left panel
Search by product title or ID
Click any image to open it in the lightbox for a closer look, where you can see full SKU details and the source image
Flag images you want to exclude from the field
Flagged images are removed from the AI image field and won't be included when you publish.
Note: The view images page always shows the most recently generated version. If you've made changes to prompts since your last generation, those changes won't appear until you regenerate.
Editing and Regenerating
You can edit prompts at any time by clicking into them from the details page. A few things to keep in mind:
You'll see a warning when opening a previously generated prompt — any changes require regeneration to take effect
After saving changes, return to the details page and click Generate to apply them
If you exit without saving, your changes are discarded
Deleting a prompt will immediately remove its associated images from the field.
Publishing AI Images to Your Feed
AI image fields can be added directly to your output feed as a treatment — no design template needed. You can also use AI image fields inside your designs like any other image column.
Aspect Ratio
When using Gemini, you can select an aspect ratio for your generated images. Options include original, 1:1, 4:5, and 9:16. The selected ratio saves with the prompt and applies during generation.
Tips for Better Results
Be specific in your prompts. Describe the setting, lighting, mood, angle, and any details you want. The more context the AI has, the better the output.
Preview before generating. Use the playground to shuffle through different products and make sure your prompt works well across your catalog — not just for one product.
Start with presets. Even if you plan to customize, presets are a great starting point. They've been tested across a variety of product types.
Use multiple prompts for variety. Different products may benefit from different treatments. Use prompt hierarchy to apply the right style to the right products.
Review and flag. After generating, take a few minutes to review your images and flag anything that doesn't look right before publishing.
FAQ
How are duplicate source images handled? If multiple SKUs share the same source image URL, the AI generates the image once and applies it to all matching SKUs.
Can I edit a field that's currently generating? No — editing is locked while generation is in progress. You can leave the page and come back when it's done.
What happens if some images fail to generate? After generation completes, you'll see a count of successful images and any failures. Failed images won't appear in the field.
Do I need a paid plan to use AI Image Generation? You can preview prompts and generate images on any plan. Publishing AI images to your output feed may require a paid plan depending on your account type.