Turn customer reviews into ad creative automatically

A cosmetics brand generates finished ad images directly from customer review screenshots. DeepSeek OCR reads the review text from the screenshot. Nano Banana Pro writes the ad creative from the most compelling part of the review. The output is a finished ad image, ready for paid social, produced from assets the brand already has.

Why are customer reviews underused as ad creative?

Customer reviews are the highest-trust content a brand can use in advertising. A real customer describing a specific result in their own words outperforms brand-written copy in almost every A/B test on paid social. But most brands have hundreds or thousands of reviews sitting in their Shopify backend or Google Business profile that never make it into an ad.

The reason is production friction. Taking a review and turning it into a finished ad image requires reading through reviews to find the compelling ones, writing copy that surfaces the best part of the review, designing the ad layout, and exporting the final asset. For a team without a dedicated designer, that process does not happen consistently.

How do brands turn review screenshots into ad images without a designer?

The approach this brand uses starts with a screenshot of the customer review and the product image. Both go into a cnaps studio pipeline. DeepSeek OCR reads the text from the review screenshot, including handwritten or styled text that other OCR tools misread. Nano Banana Pro takes the extracted review text and the product image and generates the finished ad creative: layout, copy treatment, and visual composition.

The pipeline produces a text output of the extracted review and a finished ad image in a single run. The whole process takes minutes per review.

How the cnaps studio pipeline works

This pipeline runs 7 nodes.

Step 01: Image Loader x2 (Input)
Two images load into the pipeline: the customer review screenshot and the product image. Both enter separately and are used at different stages of the pipeline.

Step 02: Text Input (Input)
A text instruction enters the pipeline specifying what kind of ad creative to produce: the target platform, the tone, the desired call to action, or any specific requirements for the output format.

Step 03: OCR using DeepSeek OCR-LT (AI Model)
DeepSeek OCR-LT reads the review text from the screenshot. This model handles styled text, handwriting, and non-standard fonts that standard OCR tools misread or fail on. The extracted review text passes to the next step as a clean string.

Step 04: Nano Banana Pro (AI Model)
Nano Banana Pro receives the extracted review text, the product image, and the ad brief text input. It identifies the most compelling claim in the review, writes ad copy around it, and generates the finished ad creative incorporating the product image. The output is a complete ad image at the specified dimensions.

Step 05: Text Viewer (Output)
The extracted review text is displayed as a text output for reference.

Step 06: Image Viewer (Output)
The finished ad image is displayed for review and download.

Why does OCR matter for review-based ad production?

Many review screenshots come from platforms that display text in custom fonts, on coloured backgrounds, or alongside UI elements that confuse standard text extraction. DeepSeek OCR-LT is designed specifically for these cases. It reads the review text accurately regardless of the visual styling of the source platform, which means the pipeline works on screenshots from Google, Trustpilot, Shopify, Instagram comments, and app store reviews without preprocessing.

Which types of reviews produce the best ad creative?

Reviews that describe a specific result in concrete terms produce the strongest ad creative. "My skin looked visibly brighter after three days" is more usable than "great product, highly recommend." The pipeline works with any review text, but the quality of the output ad copy reflects the specificity of the source review.

Long reviews with multiple claims also work well. Nano Banana Pro identifies the most compelling single claim and builds the ad around that rather than trying to include everything, which produces cleaner and more focused creative.

Try this pipeline on cnaps.ai

Cnaps.ai is a no-code visual platform for building and running multi-model AI pipelines. Fork the review to ad pipeline from the community and run it against your own customer reviews. No code required.

View and fork this pipeline on cnaps.ai

Frequently asked questions

Does the OCR step work on screenshots from any review platform?

Yes. DeepSeek OCR-LT handles styled text, custom fonts, and non-standard backgrounds accurately. Screenshots from Google reviews, Trustpilot, Shopify, Instagram comments, and app store reviews all work as input without any preprocessing. The extracted text passes cleanly to the ad generation step.

Can the pipeline process multiple reviews in one batch run?

Yes. In batch mode, cnaps studio runs the pipeline on each review screenshot sequentially, producing one ad image per review. A set of ten review screenshots produces ten ad images in a single batch job.

Does the product image change in the output ad?

No. The product image is incorporated into the ad layout by Nano Banana Pro as a visual element. The product itself is not edited or altered. The ad creative is built around the product image, not generated from it.

What ad formats and dimensions does the pipeline support?

Output dimensions are specified in the text input before the pipeline runs. Set the target dimensions to match your platform: square for Instagram feed, vertical for Stories, horizontal for display advertising. Nano Banana Pro composes the ad layout at the specified dimensions.