advertising creativity and neural networks: when and how to apply

One of the main properties of advertising is emotionality. It is through the connection with emotions that the brand is better imprinted in the viewer's memory. The emergence of neural networks has become a new stage in working with advertising creatives, because now we can go beyond our imagination and remove many organizational limitations. Let's consider how neural networks can show your product from an unusual perspective and take creativity to a new level.

When to involve artificial intelligence

The use of neural networks in marketing plays an important role for the modern specialist. Graphic neural networks are used to develop ideas for product advertising. Let's consider their benefits.

The neural network accumulates the experience of different creative people and professions: architects, artists, decorators, photographers, designers. This guarantees us the quality of the created material. In just a couple of steps, you can create a dynamic composition, create a character, develop sketches for an advertising concept, and that's not all the possibilities of neurocreators.

The speed of generating results saves time for both the design team and the customer. But successful results can be achieved by specialists who regularly interact with AI, accumulating experience and improving the quality of their skills. Therefore, our team of creatives has been studying neural networks for several months and considering ways to subsequently introduce them into projects in order to highlight a new service: neuro branding. However, mastering prompt engineering takes less time than studying the rules of design and art history, which allows us to use neural networks in projects today.

Neural networks can help when:

1. — you need to quickly develop several sketches to choose the right concept for further development;
2. — the idea is difficult to implement (for example, create a non-existent dimension);
3. — there are no comfortable conditions for content implementation (for example, team members live in different cities);
4. — you need to quickly assemble a large production team;
5. — a cinematic effect is needed.

Let's consider how to implement an idea for an advertising campaign without the need for real shooting.

Using a neural network in a creative concept

Stage 1. Creating the foundation for an idea

It is important to note that a neural network is not a panacea. It will not give you a complete idea without further manipulation. The specialist is engaged in the creative part and the selection of ideas. Successful generation of results depends on his professionalism, creativity and skill in working with artificial intelligence.

Let's assume that a browser is communicating to the audience the positioning of safe surfing on the internet using VPN and private mode. Such a browser will save user data. When developing an advertising campaign for a brand, it is important to clearly convey to potential users the positioning and main advantage of the browser: safe surfing on the internet.

The idea is based on a comparison of online actions with real life: chatting on social networks with personal communication, watching a movie online with going to the cinema. We cannot tolerate surveillance in the real world, and we should not allow it to happen on the internet either. We formulate the main idea for creatives: "This is not the time for data. This is the time for personal". After that, we plan a solution for visualizing the concept.

We have defined the direction of the project, now we will turn to neural networks: Midjourney and the built-in Adobe Firefly in Photoshop to create a creative ads.

Stage 2. Implementing a neural network

Choosing a function for a neural network

The idea of demonstrating people in life situations requires shooting in a real environment with attributes and models. A neural network can become an alternative tool to solve the task: set the scenery, select props and people, and most importantly, shoot it all. But a neural network cannot completely replace the various skills of a photographer. It will help with lighting setup, fill the frame with drama. Therefore, AI in tandem with another specialist, in this case, a photographer, gives a successful result.

But we want to showcase an impressive example of a neural network and a designer working together, where Midjourney and Adobe Firefly acted as photographers.

Unlocking creative possibilities

Sometimes, production teams require special technical equipment or extensive post-processing to create visuals. To reproduce shots that are difficult to achieve using basic tools, neural networks come to the rescue.

Let's consider an example of a scene where a heroine is being followed in a changing room during shopping. To achieve a natural look, a wide-angle lens and specific lighting are essential, but these are just a couple of conditions for successful implementation. However, a neural network can generate several identical frames. Afterward, a designer can fine-tune them in an editor. In the prompt, you should specify "lens 5 mm" or "fish eye" to obtain the desired wide-angle lens effect.

In this case, we've shown that neural networks can overcome various limitations hindering the realization of a designer's creative vision.

Continuing to work on ready-made frames

It's best to provide neural networks with references for the specific frame you need. Otherwise, they produce simple and mundane results, as they don't always understand the right perspective or composition. We can make neural networks more effective by providing example similar frames that they can refer to during image generation. This saves time and yields more accurate results.

Frames from the world of cinema are ideal as references, as they have been crafted by professionals from various fields: cinematographers, costume designers, and artists.

But let's complicate the process: We'll take a photo from the poster of the series "Patrick Melrose" featuring Benedict Cumberbatch and use it as an example for the neural network, but modify it to suit our specific purpose. This might spark a debate regarding the legality of such actions. Let's look at two illustrations, before and after processing, to settle this dispute.

We used someone else's work as a reference for the neural network and modified it almost entirely. We created a new creative product by globally reworking the reference. From the original frame, we only took the angle in the bathroom, which is unlikely to be patented by anyone.

Let's look at the sequence of transformation stages from the reference to the final image for a communication strategy.

Choosing a clear prompt for the neural network

Let's showcase a different scenario: A tough bodybuilder is moved to tears watching a melodrama. It's unlikely many people would want to be seen, let alone filmed, under such circumstances. The concept of privacy is relevant here.

Creating the creative, starting with reference selection: a bodybuilder and the angle in a movie theater. The neural network has learned to reproduce human emotions on request. For instance, for tears, which we need in the final image, we don't need to search for a reference.

It's not necessary to provide the neural network with a large number of images, as it is very receptive to them. Pay more attention to the quality of the selected references rather than their quantity, which can disorient the network. Here's what we achieved:

The first one looks like this: brutal bearded man crying in the empty cinema, eating popcorn, lens 20 mm, shocked face, photorealism, detailed facial, global illumination..

The neural network generated an illustration, but not exactly the one we needed. "Crying" in English has two meanings: "weeping" and "shouting." While the neural network fulfilled the request, we needed the first meaning of the word. Therefore, we used the phrase "laments with tears" to achieve the desired facial expression. We select various analogies and associations when writing the prompt. It's worth noting that due to language differences, literal descriptions are required. Here is the result, which we will finalize with post-processing:

To see the difference between the prompt, pay attention to the wording of the final: «holds busket, big busket of popcorn, red seats, brutal powerlifter man in t-shirt laments with tears in the empty cinema, busket of popcorn in front, lens 20 mm, photorealism, detailed facial, global illumination».

Stage 3. Refining the image to its final version

At the beginning, we mentioned the tandem of neural networks and designers. At this stage, we refine everything the neural network has created, choosing the best options:

Finalizing the image:

1. — Retouching: We eliminate visual flaws and correct lighting and color errors;
2. — Adding Elements: We convey the ambiance and dramaturgy of the creative;
3. — Adjusting Anatomy: We correct the shaping of fingers and other body parts, as neural networks struggle with fine details;
4. — Adding Design Elements: We unify the resulting image, add typography, brand elements to achieve visual unity;

Returning to the previous creative with the bodybuilder, we note that the neural network failed to generate tears, so we added them to create an emotional effect. This further proves that neural networks are merely tools that produce the desired result only with active participation from the creator.

The creative with the actor has a lot of empty space. We want to increase the contrast between the hero's brutality and the private nature of the situation. Using a neural network, we create rubber ducks that he lies in the bathtub with or even plays with punk music. We add bold tattoos through retouching, which will add more brutality. As a result, the creator thinks through the idea and its amplification through the confrontation of brutality and naivety, while the neural network helps find the best creative solution and implement it.

A new reality

Everything described above is actively used in advertising today. Using neuro branding, we can create visuals that stand out among ads created using traditional production methods. Neural networks still cannot replace creators, photographers, and designers, but now there's a choice of how to achieve a spectacular result.

Let's take a look at the completed visuals used in the advertising campaign. The final layouts solve the task set above.