If you let a tool do all the work, you become the tool.

With any new technology, there is a honeymoon period where the world is convinced it’s the ultimate solution to any problem. As time progresses, it has its failures, and the pendulum swings back sharply. The technology is suddenly vilified as incapable of solving anything. Then, ultimately, it finds its niche and is used to solve problems in the specific use cases that it’s best suited for.

When it comes to solving creative problems, to be of true use, any technology must be treated as what it really is—a tool. Such is the case with AI, and image generation specifically. A tool is only as good (or bad) as the skill of the person using it.

The right tool for the right job.

An obvious use for image gen is in the case of quick-hit content like social media posts, where new content is expected daily, but budgets seldom allow for the full craft in photoshoots, or the hand-illustration needed for stunning and relevant visuals. However, when everyone else is using tools like Midjourney to create their content based on the same LLMs and generic prompts, there is no ability for market differentiation or a unique brand voice to shine through.

As such, it was important that we used AI image gen as just another tool in the toolbox rather than an end-all-be-all way to create social media content. This meant that our own creative vision and the voice of the brand could come through with an approach that combined image generation with iterative functions and hand-crafted design. At the end of the day, controlling the creative process was paramount for work to hit the mark.

A hammer for every nail.

The first step in this process was the tried-and-true brainstorm to think of the types of visuals we wanted, keeping our eyes on audience engagement and our brand goals. Specialty apples are a crowded space where having an identifiable and unique visual approach is the best way to stand out.

From there, it was time to bake the bricks that would go into the building. This included custom photography to capture the look of a real SweeTango, as well as some stock photography, and finally image generation. As each LLM and image generation tool has its own strengths and weaknesses, we used multiple tools like Midjourney, Flux LoRA, Adobe Firefly, and Photoshop generative fill to achieve what matched our vision and design tastes.

Importantly, we kept in mind the old wisdom that only a fool accepts what they are first offered. This meant iterating on those generated images to refine them. To do so, we prompted Midjourney to describe the semi-successful images in a way that AI itself would understand best. Then, we fed that generated prompt back into all the AI image gen tools to improve on the initial results.

Finally, at the end of the chain, our design studio took these resources and combined them by hand into Photoshop, refining and fixing issues for ultimate creative control. This is the crucial step that transforms content from a C- to an A+. It’s a step often crucially overlooked.

Handmade creative that creates results.

In addition to building our organic social media presence, paid media told the true story of the value of creative directed output with a 175% increase in ad clickthrough and 162% increase in website visits. Most importantly, when backed by media spend, the delicious imagery drove over 18,000 visits (a 104% increase) to SweeTango’s store finder app.

Use the tools you have, but don’t let the tools use you.

It’s important to never get too high or too low on any technology. While AI isn’t the ultimate solution to hitting a budget and getting eyeballs, it can be a part of a smarter way to approach creative that works. More follows, more likes, and more sales all comes from more creative thought put into your marketing.