How to Use Fine-Tuned Stable Diffusion Models to Generate Riveting Advertising Creatives

Introduction

We already love Stable Diffusion – but now we have Civitai, which is a place where those experimenting with Stable Diffusion to develop their own models have a platform to share their work with other users and AI enthusiasts. On Civitai you can post your own Stable Diffusion-based AI image model for other users to discover, as well as you can post AI-generated photos for consumers to browse through and engage with.

As the community around Stable Diffusion grew, people were learning how to do different things with the model through prompting or by adding new concepts into the model, like using images of themselves to make AI-generated selfies, for example. But these were being posted around the web in places like Reddit or Discord, not in a centralized community. Thus was born Civitai, an online community that organized this content in a way where people could come back and find it later. Over time, it became like an image-centric Hugging Face of sorts.

How to Use Civitai

To use Civitai, users can upload a series of images that represent the style those images represent, then choose the base model — like the standard model, anime model, or more realistic model, and so on. Then you’ll be able to generate your own images using the style you have captured. Civitai can be a powerful free resource for advertisers and marketers to generate high-grade ad creatives that are both quirky and realistic. It's like having an artistic genie at your disposal, ready to bring your vision to life. Whether you're in need of captivating headshots or eye-catching product photos, Civitai can be your creative ally.

Picture this: You're a budding photographer aiming to capture the essence of a client's unique personality in their headshot. Instead of countless hours spent behind the camera and editing desk, Civitai allows you to generate a range of headshot options effortlessly. Simply fine-tune your diffusion model on the individual's facial features, preferences, and unique style, and voila! A gallery of professionally rendered headshots that are as unique as your client's personality will be generated.

For businesses seeking to showcase their products, Civitai can be a game-changer. You can do away with expensive and time-consuming photoshoots. With custom diffusion models, you can generate high-quality product images that highlight every detail, from intricate designs to flawless finishes. Imagine the convenience of instantly creating an entire catalog of product photos, ready to dazzle potential customers. You can fine-tune your diffusion model to generate art that resonates with your creative vision, or craft engaging visuals for your content.

Generating Ad Images with Civitai

What if you want to create some images of the Hyundai Verna as part of an advertising promo? Here’s what you need to do.

First log in to Civitai.com.

‍

Then click on ‘Models’ to search for Stable Diffusion Models that have been fine-tuned on car images. We found a model called Lada Niva, which seemed like it could do the job. Let’s try to generate some images with this model.

Click on the ‘Create’ button, and let’s experiment with the prompts:

‍Prompt 1: black hyundai verna, inside a studio, white background, shadow, light does not reflect off the car

There is an option for providing a Negative Prompt, which means what we don’t want the image to be. We’ll use the following:

‍Negative Prompt: (cinematic-bloom:1.38), (haze:1.35), (fused-knees:1.32), (fused-legs:1.31), (monochrome, grayscale:1.3), (unfocused,blurry:1.3), (DOF:1.2),(tilt focus:1.2), (nude breasts, nipples:1.1), (wide-shot:0.5), (simple surroundings:1.39), (empty background:1.395), (worst quality:1.39) (low quality:1.39) (normal quality:1.3), bad hands, deformed fingers, author signature, text, watermark, (cropped::head:1.35), (close shot:1.3), (key lighting:1.2), (brightly lit:1.3), easynegative, bad-hands-5

That looks good! We tried a white version of the same prompt and got this:

Prompt 2: red hyundai verna in an open field, evening scene

‍Negative Prompt: Same as before

Prompt 3: Hyundai Verna, front facing, headlights on, dark background, low light, no light, highly detailed, highly accurate, awe-inspiring, dramatic lighting

Prompt 4: Hyundai Verna, day scene, on dirt road, racing, dirt on the car body, cinematic action, moving car, no reflection off the car body, photography, realistic, intricate detail, details, crystal clear, perfect face, ultra sharp

This trial run exemplifies Civitai’s potential to create tailored professional images for diverse purposes. Using the Lada Niva model, the platform successfully generated captivating images of the Hyundai Verna, meeting specific criteria such as studio shots with a white background and detailed prompts for various scenarios.

Next Steps

For fun and games, Civit works well. However, for serious business use-cases, you should ideally work on a model fine-tuned specifically for your use-case. You can do this through platforms like Astria.ai, which offer an easy and streamlined approach to fine-tune Stable Diffusion models.

Alternatively, you could attempt to fine-tune stable diffusion yourself, as we have already shown in a blog article. Fine-tuning model for your specific domain gives you far more control than you would get from publicly released models, and is the ideal approach for businesses who are looking to stay on brand where quality and accuracy is of extreme importance.

In conclusion, the possibilities are endless. If you are an artist, you can use the tool to create art using text prompts. If you are a teacher, you can use it to create creative visuals to teach students and help them understand concepts faster. If you are a designer,  you can use the tool to create designs for any D2C product. Try it for yourself and witness the results!

About Superteams.ai

Superteams.ai connects top AI talent with companies seeking accelerated product and content development. These "Superteamers" offer individual or team-based solutions for projects involving cutting-edge technologies like LLMs, image synthesis, audio or voice synthesis, and other cutting-edge open-source AI solutions. With over 500 AI researchers and developers, Superteams has facilitated diverse projects like 3D e-commerce model generation, advertising creative generation, enterprise-grade RAG pipelines, geospatial applications, and more. Focusing on talent from India and the global South, Superteams offers competitive solutions for companies worldwide. To explore partnership opportunities, please write to founders@superteams.ai or visit this link.