Case study: Testdrive MG - an AI-campaign
This project was curated by deBottomline. The experiment is a result of not using traditional photography.
Sense
Framing Context
Research question: create an AI workflow for the MG brand + a developed final product
MG was looking for an efficient workflow to quickly, easily, truthfully and affordably generate images using AI software to advertise their brand in the European market where the budget is limited.An example campaign had to be created to illustrate this solution in the form of a video production or animation for social media, aimed at creating awareness for MG among people looking for a new car. The core message: ‘affordability meets technology: it’s a match!’
Framing Goals & Intent
Design question 01: Why are people attracted to a particular car model?
My research showed that each type of car model has its own specific target audience, based on the needs of potential customers.
Design question 02: Which AI tools and visuals will I use to create the desired workflow?
To meet the requirements, I researched AI technologies for different stages of the production process: Stable Diffusion to train a dataset, Runway and Kaiber for animation, visuals generated with the Luma AI app, concept art with Midjourney, and AI voice-over and scripting via ChatGPT.
Sample
Exploring Ideas
I came up with the idea to work around a test drive campaign to create personalized content for each target audience. This allowed me to convey MG’s message in a unique way.
01 Create moodboards per car model
02 Develop a campaign theme
03 Experiment with various AI tools
04 Storyboard
Shape
Prototype
01 Clean up 3D model
02 Build different datasets based on visual material and train them via AI
03 Create visual effects/animation using AI
04 Test social media format
Stage
Testing
The final product was tested in an MG showroom. People were surprised that it was all made using AI and thought it was a fun idea.
Integrating Result
The main finding from the research was that understanding the needs and expectations is essential to effectively reach the target audience. The final product was designed so that users can generate images of the car models using custom prompts in AI software, enabling reuse of the dataset for future projects.
Sense
Framing Context
The project explores the use of generative AI in pre- and post-production to create an awareness campaign for the MG brand. The focus is on realistic AI-generated visuals with a limited budget, especially for the MG4 model, targeted at West-European audiences.
Framing Goals & Intent
Research and design questions:
– How can MG’s core values (comfort, spaciousness, affordability, innovation) be communicated visually?
– Which AI tools are best suited to generate realistic car visuals (e.g., Stable Diffusion, Midjourney, Pika Labs, RunwayML)?
Sample
Exploring Ideas
Three concepts were explored:
1. Family-oriented MG4 for everyday scenarios
2. ‘MG4 as a friend’ in surreal or fun settings (e.g., car on carousel)
3. ‘Back to the Roots’ – placing MG4 in symbolic British locations such as Abbey Road, pubs, Buckingham Palace, tea time and golf scenes
Prototype
Multiple AI tools were tested:
– Stable Diffusion with custom 3D scans and embedding via textual inversion
– Polycam and LumaAI for photogrammetry
– RunwayML and Pika Labs for animation
Initial test renders showed distortions and unrealistic outputs. Prompt crafting, negative prompts, and technical parameter tuning were refined.
Shape
Testing & Training
Stable Diffusion and RunwayML were used to generate realistic and animated images. Several test renderings were evaluated. A 3D model of the MG4 was successfully scanned and cleaned for training.
Stage
Result
Despite technical limitations (GPU, render accuracy), a functional AI workflow was created. Campaign themes like ‘MG4 Skiing Through 2024’ and ‘MG4 Your Friend’ were developed, and deliverables were handed to deBottomline, including a trained model and example prompts. Further refinement is recommended.
