AI, 3D and Product Integrity: Rethinking the Visual Production Workflow
How we combined AI and 3D without compromising product integrity.
What Happens When the Product Comes Second?
In our previous post, Pipeline, Process and the Communicative Quality of Visual Content, we explored how the structure of a visual production pipeline can influence the final image.
This project took that question one step further: what happens when you define the visual direction before deciding which motorcycle will appear in the scene?
The brief was deliberately incomplete. No model, color or product identity — only one question: what should the image communicate?
The answer had to come before any software, subject or scene was chosen.
*Not official Visual – Internal project
The Editorial Phase: Before the Subject
For several weeks, we developed the composition, lighting, atmosphere, colour palette and emotional tone without knowing which motorcycle would eventually appear.
Research and references came from well beyond the automotive and motorcycle industries. The challenge was to make aesthetic decisions without relying on a specific product as an anchor.
Only then did we introduce rapid visualisation tools such as Midjourney, Adobe Firefly and other AI image generators.
Not to define the direction, but to test it visually.
Without a clear direction, these tools produce an abundance of plausible images. Used afterwards, they become a form of rapid visual sketching.
Selecting the right results still requires judgement: understanding what works, what remains coherent and what moves the project away from its intended tone.
That responsibility remains human.
Why Run a Project Without a Client?
In established pipelines, marketing teams often define the editorial direction while production teams focus on execution. This is efficient, but it can create distance between the two.
This internal project was designed to reconnect them.
The goal was not simply to produce images, but to bring editorial decisions back to the centre of the process: composition, atmosphere, tone and emotional tension.
Working without a client brief or deadline also allowed us to document decisions, test assumptions and revise the work when necessary.
The Product Never Goes Through AI. Ever.
The product’s master 3D model — including its geometry, proportions, materials and defining details — remains outside the generative layer.
We do not ask AI to reinterpret or reconstruct it.
This is both a technical and a strategic choice. A motorcycle’s Digital Twin is a controlled, accurate representation of the product, capable of supporting different configurations, updates and formats over time.
The goal is not simply to create a plausible image, but to preserve a controlled source of truth and its IP.
This is especially important when working with unreleased products, confidential specifications or future configurations. It is not enough to know whether a tool can process such material; you also need to understand how the data is handled and what confidentiality guarantees apply.
In our workflow, the product’s master geometry remains outside the generative process by design.
Knowing where technology adds value — and where it could reduce control or create unnecessary risk — is part of being a reliable visual partner.
Transparency Starts With the Workflow
The EU AI Act introduces a risk-based framework for AI systems, including transparency obligations for certain AI-generated or manipulated content.
The European Commission’s AI Act page provides an overview of the regulation and its implementation, while further information on transparency obligations explains the responsibilities of providers and deployers.
The exact requirements depend on the system, use case and role of the organisation involved.
For us, transparency begins before the final image. It is built into the workflow: deciding which elements can enter the generative layer, which must remain linked to controlled assets and where human responsibility must remain.
We adopted this principle primarily to protect product integrity.
AI can expand the creative space without becoming the source of truth for the product.
The Technical Challenges: Building What Did Not Exist as an Asset
Once the direction was defined, production introduced a series of practical challenges.
The rider was essential to the scene, but no existing asset met the required quality. We therefore built it from scratch, developing the clothing and protective details element by element.
Environmental effects, from smoke to sand, were tested with different tools and techniques to balance realism with control.
Finally, the motorbike – the main focus of the campaign, though not of our training: without altering the geometry of the digital twin, all the materials and details required to integrate it into the visual concept were reviewed and adapted.
Post-Production: Closing the Loop
In the final stage,3D, compositing and AI tools were used to refine the images, improve consistency and resolve specific details.
Tool selection was always based on the material being handled and the level of control required.
This is particularly important in client work, where product assets, confidential information and unreleased content require careful consideration of not only what a tool can do, but also what we are giving it access to.
That is not a limitation. It is part of the method.
What We Took Away From It
The main result of the project was a clearer understanding of where the most important decisions are made.
They begin before the technical phase, with composition, tone, emotional tension and the relationship between product and context.
No tool can define these elements automatically.
For us, responsible AI use means knowing what to delegate, what to keep under direct control and who remains accountable for the final result.
Working without a client brief or production pressure allowed us to test this approach in depth. The value it brings to client work is not a list of tools, but the ability to build processes that preserve creative freedom, product integrity and long-term control.
If you are preparing a launch or managing visual assets for a complex product over time, one of the best ways to evaluate a partner is to understand which questions they ask before they begin.
I believe great communication begins with listening. With a background in design and five years leading In2real’s communication efforts, I work with our team to help brands connect with their audiences through clear strategy, visual quality, and digital storytelling.