Generating creative

Briefs and media plans often call for a huge quantity of assets—a complex matrix of permutations for imagery, copy, calls to action, and more. However, this large volume often masks a limited exploration of truly diverse creative concepts. Think of it like this: you might have thousands of variations, but they're all built on a small set of core ideas. While certain elements, like copy or color, can be easily automated, the fundamental building blocks of creative—imagery, video, and design—have historically been constrained by the time-intensive nature of manual production.

Generative AI solves this challenge. By integrating these new tools with our existing automation systems, we can now produce Creative at scale, allowing for more responsive, more engaging, and more effective campaigns.

A trained stable diffusion model can produce infinite AI generated product images.

Developing a new pipeline

Off-the shelf Generative AI products like MidJourney and Adobe Firefly are incredible tools for exploring ideas and creating one-off images and videos. But in the context of large scale campaign production, we need more control, consistency, and image quality.

To accomplish this, we are developing custom GenAI workflows - fine-tuning existing diffusion models with custom LoRAs. Trained locally on existing brand/product imagery, these refinements create guardrails for the AI, ensuring brand-alignment and consistency between generations.

The pipeline combines new image generation processes with existing automated post-processing tools.

With the model trained and tuned, we begin exploring concepts and themes through text prompting. With a generation time of seconds per CPU, we are free to iterate at the speed of inspiration - evaluating thousands of images, working to curate only the strongest ideas.

Once selects are made and refined, the remainder of the automation process is relatively traditional. We dictate dimensions and resolutions, adjust composition, fill in gaps with in-painting, apply text, graphics, and color correction - then export.

Once the model is trained and the templates designed, we can produce content as fast as audiences can consume it.

Flexibility for modern media

Automation fundamentally transforms the mechanics of creative production. Increased content volume translates to a vast, dynamic library of assets. But this isn't just about quantity; it's about having the right variation instantly available, tailored to specific needs and contexts.

A production process that enables near-instantaneous generation of new iterations allows for immediate responsiveness to evolving creative briefs or unexpected campaign needs.

Automated image generation and post-processing can produce content faster than you can consume it.

For brands with high-volume content needs, in rapidly changing markets, with highly segmented audiences, or operating across culturally diverse landscapes – automated creative production is here and it’s immediately an indispensable tool.

From Lego to Van Gogh, models can be tuned to create imagery in any art style.

Images are first generated at lower resolutions to screen for content, then upscaled with another generative pass to fill in detail and render for higher resolutions.

Image composition and dimensions are customizable based on media platform.

These images and use cases are provided solely for illustrative purposes. Crocs, Inc. is not affiliated with Hook, its technology, its processes, or this content.