Is AI the ultimate solution to our aspirations of conducting product photography independently, without the need for a photographer, stylist, props, elaborate lighting setups, and the significant costs associated with them?
Recently, I’ve delved into the realm of the latest AI applications, recognizing that by the time you read this, there may already be a plethora of new options (and upgrades to existing apps) available. My objective? To take a look at their capabilities and see what they bring to the table.
Note: At the end of this post, I’ll give you a list of the apps I’ve tried.
My Criteria
My main goal for evaluating these apps revolves around (what should be) the seamless process of capturing a product photo, uploading it into an AI-powered platform, generating a specific background, and finally downloading the resulting image in high resolution (for print), ideally at 300dpi at a full page size of approximately 9'' x 12''.
Placing Product Photos and Background Generation
The initial hurdle arises when attempting to insert a product image into an AI application and subsequently generating a fitting background. While most apps offer this functionality, the challenge lies in preserving the integrity of the product image amidst the AI’s background generation process. In my trials, the product image often underwent alterations, ranging from subtle to significant. Ensuring that the product image remains pristine and untouched during background generation seems to be a complex task, yet one that must be (if we are to use this technology effectively) achievable.
Take a look at the images above. On the left are the original product photos, and on the right are the AI-generated versions.
• • •
[Top] On the audio rack image (created using Flair.ai), look at the bottoms of the legs (why are they suddenly gold?); the shape of the turntable; the connection points where the top shelf is attached to the legs… Some of these differences are subtle but if you’re selling that audio equipment rack, you need it to be a 100% accurate representation of the product.
[Bottom] The image of the hot chocolate tin (created using Pebblely) has a simpler shape and I think the AI generated image is much more acceptable. Pebblely also offers editing controls, allowing you to edit the product image itself, separate from the AI generated background. However, the tin still needs a bit of editing as it does not look 100% believable that it is sitting on that countertop.
• • •
Resolution Woes
Many of the apps I tested max out at a resolution of 2048 x 2048 pixels (28.4'' square) at 72dpi. However, when aiming for print quality at 300dpi, the image dimensions shrink to just 6.8'' square, which may suffice for certain applications (and is totally fine for online/website/social media usage) but falls short for print formats such as full-page images or catalog covers.
Attention to Detail
In the age of AI, we've grown accustomed to the occasional imperfections in generated images. Whether it’s the notorious misrepresentation of hands or subtler inconsistencies in backgrounds, surrounding props, or textures, the results can often appear unnatural or off-putting. Such discrepancies, though sometimes subtle, can adversely affect the visual appeal of marketing materials, potentially undermining brand credibility.
Final Thoughts
While there's undoubtedly promise in AI development, it’s evident that there’s still ground to cover for AI developers. Although I’m optimistic about future advancements, particularly in catering to complex product shapes, the current landscape may leave some users dissatisfied with the results.
Have you explored AI applications for product photography? Have you discovered any that meet your needs without compromising quality? If so, I'm eager to hear about your experiences and insights.
• • •
P.S. Here are the apps I’ve experimented with:
Comments