The new ChatGPT 4o update makes AI image generation more accessible and impressive than ever. What previously required a lot of prompt engineering, post-processing, and trial & error can now, in many cases, be achieved with a single instruction. As soon as this functionality becomes available through the API, we will immediately capitalize on it. Below we show four concrete applications that we are already testing and that can soon be fully automated.
Example 1: Generating realistic backgrounds for freestanding products
For our client Acaza, we placed a side table that originally stood on a white background in a modern living room. What makes this new technology unique:
- The background is not simply placed "behind" the object, but perfectly matches the product in terms of perspective, lighting, and style.
- The AI has even subtly adjusted the angle of the object, as if the photo was retaken.
- Previous tools struggled with incorrect proportions or mismatched styles. That problem now seems definitively solved.
Input:
Output:
Used prompt: Place this side table in a modern Dutch setting. It should look as if it's a professional product photo, made in a carefully composed interior in a studio. Ensure that the lighting, shadows, and colors of the model match the white studio background and the existing side table photo. The focus should completely remain on the side table, so few distracting elements and accessories. The style must match a high-end furniture photo.
Example 2: Generating a suitable model for existing clothing
For a work pant from Fristads, we wanted to generate a model that exactly wears this garment. The AI has:
- Created a realistic-looking model with correct anatomy (no more six fingers!).
- Correctly and recognizably applied the exact garment, without deviating textures or wrong fit.
- Generated a natural pose and expression, as if it were a real photoshoot.
This was previously hardly achievable without custom 3D modeling or extensive Photoshop workflows.
Input:
Output:
Used prompt: Add a male model to this studio photo of a pair of work shorts. The model should wear the shorts as if it is a professional product photo, with a neutral expression, standing slightly upright in a natural pose. The product is workwear, aimed at the construction industry, so he can be a rugged-looking but well-groomed man. Ensure that the lighting, shadows, and colors of the model match the white studio background and the existing shorts photo. The focus should remain entirely on the shorts, so no distracting elements, accessories, or backgrounds. The style should match a high-end fashion PDP photo.
Example 3: generating people into an existing background with a cut-out model
We had a model generated in a carnival outfit and then added people in the background in a matching atmosphere. The result:
- The model remains sharp and realistic, while the generated background people naturally contribute to the atmosphere.
- The colors and lighting are perfectly matched.
- The AI understands context: no people in suits at an Oktoberfest, but a partying carnival crowd in matching outfits.
Input:
Output:
Used prompt: Add a background to this studio photo of a man in carnival clothing. It should look as if this photo was taken in a city center where a carnival is taking place. Ensure that the lighting, shadows, and colors of the model match the white studio background and the existing shorts photo. The focus should remain entirely on the existing man in the carnival clothing, so do not add other people, distracting elements, or accessories. The style should match a high-end fashion PDP photo.
Example 4: generating category hero images based on product photos
For A4H-Tech.nl, we took three different exhaust tips and had a category header generated from them:
- The AI rotates the products to a logical perspective and scales them proportionally.
- A background of an auto workshop is generated, perfectly fitting the subject.
- The composition is directly usable as a hero image on a Magento category page.
Input (3 separate images):
Output:
Used prompt: Here are 3 photos of exhaust tips for Honda cars in the tuning market. I want you to generate a hero image for the product category "Exhaust Tips" using these photos. The background can be, for example, an auto workshop where these products are ready to be installed. The focus should remain entirely on the existing exhaust tips, so do not add other products, people, distracting elements, or accessories. The style should match a high-end PDP photo.
The hero image should have a 16:9 ratio. Do not add any texts to the images yourself.
Of all the examples, this offers the most potential: many customers have good product and atmosphere photos, but lack strong, composite images for their category pages. Once the API becomes available, we will develop a Magento extension to automatically generate category images for this purpose.
With Sora, even a short video can be made from this:
What's coming up?
Once the API is live, we can fully integrate these applications into our processes:
- Automatically generated category images for Magento webshops.
- Smart generation of banners, ads, and social media visuals.
- Generating product variant images (color, context, target audience) without photo shoots.
AI image generation is no longer a gimmick. The combination of ease of use, realism, and control opens up a whole new spectrum of visual possibilities.
Curious about possibilities for your webshop? Contact us!