Creating a Brand With AI

A few months ago, I started the process of building a brand for a startup. With a background in product management and fine art I took on the challenge because in a startup environment being a multi hat wearer is kinda a necessity. It’s been more than a few years since I managed a website development project or marketing campaign, and the process has changed a ton. I have overseen a lot of teams and understand the tools, or at least I thought I did. With the easy accessibility of AI image and video generation tools I set out on a quest to explore what’s available. It’s so far been an interesting, cool and aggravating ride.

I’ve been messing around with generative AI and diffusion models for a while now and have really enjoyed watching the tools and workflows expand and evolve. They’ve come a long way, fast. But the workflow to create brand consistency is not straightforward. As it is currently you still need multiple tools to get a desired aesthetic and keep it consistent across assets. Depending on the tool, model, and prompt you can get wildly different outputs. I’m curious how models respond to the same prompt, evaluating the output for consistency, interpretation and creative control.

Let’s start with creating something pretty simple: a cat, walking through a forest. The web is full of cat imagery so I expect it to be pretty consistent. We already have an established brand color palette. Let’s see what we get from OpenAI, Gemini, Adobe Firefly and a couple models using Runway UI, also note the processing times. In my experience there’s a huge difference between generating through an application, via API or directly. But that’s for another time. For now we’ll generate some cat images, pick our favorite and use it as a reference for video clip creation.

Prompt:

create an image of a single blue cat character walking through the woods in cinematic realism facing the camera. Body type: lean adult domestic British shorthair Coat: light blue color throughout with no other markings. Eyes: dark green Tail: long, single, naturally proportioned neutral expression entire body within frame Background is a wooded area in the brand aestetic palette: Premium colorful brand-world visual language: modern geometric editorial illustration with mid-century influence translated into polished cinematic dimensionality; large readable shapes, elegant curves and diagonals, calm negative space, deep teal shadows, warm sand and clay neutrals, saturated orange and turquoise accents, selective yellow, magenta, and electric blue; tactile woven textiles, convincing glass, metal, water, and wet-road reflections; integrated inviting technology, restrained atmospheric depth, coherent geography, stable light direction, emotionally grounded surrealism, hopeful and curious rather than dystopian. The image does not contain any visible text. Output images below, which one do you like the best?

Pretty cool results, varied but all decent.

• Firefly is the least accurate. Is that a walkway light post next to the cat? But it had the quickest processing time.
• I really like the aesthetic of the ideogram images: the colors, woodland background, etc. But it is the only one that doesn’t have any wet road reflections, and the cat isn’t really blue.
• The Chatgpt one is cool, more cinematic than the rest and I like the blue color. Still, it’s a bit off as the cat should be more cinematic with a background that has an illustrative quality.
• Gemini does a better job dealing with the prompt and even incorporates the tactile woven textiles detail into the image.
• Nano Banana 2 does a very nice job with the prompt and incorporating the details into the image. It has the blue cat, the wet-road reflections, integrated inviting technology, the correct color palette and mood. It fits my brand the best:

Now, let’s take this one as a reference and use it to create a 5sec video, 720p, 24fps, 16:9. We’ll use Firefly video, Runway for several, Chatgpt for comfy api access, and Gemini.

Modified prompt to include camera angle:

A full shot single blue cat character walking through the woods in cinematic realism facing the camera. The cat approaches the camera Body type: lean adult domestic British shorthair Coat: light blue color throughout with no other markings. Eyes: dark green Tail: long, single, naturally proportioned neutral expression entire body within frame Background is a wooded area in the brand aesthetic palette: Premium colorful brand-world visual language: modern geometric editorial illustration with mid-century influence translated into polished cinematic dimensionality; large readable shapes, elegant curves and diagonals, calm negative space, deep teal shadows, warm sand and clay neutrals, saturated orange and turquoise accents, selective yellow, magenta, and electric blue; tactile woven textiles, convincing glass, metal, water, and wet-road reflections; integrated inviting technology, restrained atmospheric depth, coherent geography, stable light direction, emotionally grounded surrealism, hopeful and curious rather than dystopian. The image does not contain any visible text.

Wan 2.2 via comfy, processing time 4m22s

Flux 3 via Runway, processing time 11m24s

Seedance 2.5 via runway, processing time 9m46s

gemini omni, processing time 2m10s

Firefly video, processing time 2m54s

Kling 3_0 Turbo via Runway, processing time 14m10s

How did the models do? All the models did pretty good, and cats are a good easy model test because there’s sooo many reference images on the internet, I would not be surprised if the models ran on cats. But it’s the blue, the background details and the walk that count the most.

• Flux and Seedance do a slow walk approach and had surprisingly similar darker background aesthetics.
• Firefly’s tail movement I really like.
• Gemini created a longer clip which was their default, I guess. No complains and the purr in the audio is a nice touch.
• Wan also had pretty nice details and I like the walk, for me the pace seems more realistic.
• Kling was the only one that drifted significantly from the prompt with a 360-degree camera rotation and the cat doing a complete turn. I thought it was a cool effect but not what was requested.

So, which one is your favorite? For me it’s tough. I think it’s a toss up between Wan, Gemini and Firefly. Which would you choose?