One Race Car: Two Design Processes

The Experiment

Artificial intelligence is changing almost every aspect of the creative industries, and automotive design is no exception. Generative AI can produce images in seconds that might once have taken a designer hours, days or even weeks to develop.

The automotive design industry is already beginning to grapple with what AI might mean for the profession. Designers are experimenting with generative AI to explore ideas, rapidly producing alternatives and pushing concepts in directions that might otherwise take considerably longer to investigate.

In this article, we’re not questioning whether AI can generate a convincing image of a car — it clearly can. The more important question is: what happens when an experienced designer uses that technology as part of their own creative process?

Rather than trying to answer that question with a simple human-versus-AI competition, TorqArt decided to try something slightly different.

We gave a professional car designer a classic car and a brief — and asked him to tackle it twice.

The Brief

“Create a modern interpretation of the Porsche 917”

Design a contemporary endurance racing car inspired by the character, proportions and spirit of the Porsche 917.

The result should feel unmistakably connected to the original 917, while being a new design rather than a simple restyling. Consider how its proportions, aerodynamics, cockpit, packaging and details might evolve for a modern endurance racing environment.

The same brief will be approached in two ways:

01-Artist-led: The designer develops the car through their normal design process, from initial hand-drawn sketches and concepts through to the final design. Final renders may combine traditional techniques with AI-assisted tools.

02-AI-assisted: The designer will use an AI-assisted approach to explore and develop a design from the same brief, documenting the prompts, iterations and revisions through to the final result.

Objective: The objective is not to determine a winner, but to compare the two creative processes and understand what each approach can — and cannot — contribute to automotive design.

We’re essentially asking… what does an experienced designer see that AI doesn’t? And what can AI help a designer accomplish that they might not have considered?

And perhaps most importantly… does using AI actually change the way a designer thinks?

The Challenger

For this experiment, we wanted someone who understood both the discipline of professional automotive design and the importance of interpretation, which led us to Tony Hunter.

Tony is a Scottish automotive designer with a career spanning several decades, largely specialising in automotive interior design. His career has included prominent brands such as Renault, BMW, MINI, Jaguar Land Rover, Ford, Volvo, Tata and McLaren.

But it was some of Tony's personal projects that made him particularly well suited to this experiment.

One of Tony’s most remarkable examples is his James Bond “The Locomotive” Bentley.

Inspired by Ian Fleming's original descriptions of Bond's Bentley, Tony was determined to turn Fleming's fictional car into a physical reality. This was no small feat, considering no such vehicle — or even images of one — actually existed. Tony therefore had to interpret Fleming's descriptions and apply his own ideas about how the car might look, ultimately turning a fictional automobile into a stunning physical reality.

"The Locomotive" Bentley

Another example can be seen in one of Tony's more recent projects, the car he calls the “Emiura”, which combines elements of two of the most celebrated sports cars ever designed: the Lamborghini Miura and the Jaguar E-Type.

When I first saw Tony’s concept images, my immediate reaction was that this design deserved to become a physical car. It had all the right ingredients for a beautiful, modern-day, retro-inspired sports car — no fuss, no aggressive pointy bits, just a beautifully proportioned design with gorgeous classic curves.

Rather than simply creating a modernised version of either car, Tony produced something that feels like its own design, with the visual drama and proportions associated with the great Italian and British sports cars of the past, but expressed through a contemporary interpretation.

"Emiura" Concept

Tony understands how to take an established visual language and reinterpret it without simply copying it. He understands proportion, surface, detail and, perhaps most importantly — restraint.

The Porsche 917

Few racing cars have a visual identity as strong as the Porsche 917.

Porsche 917K

The challenge with a car like the 917 is that, beyond its remarkably identifiable tail, its identity isn't really contained in any single detail. It is the combination of its overall shape, low stance, proportions, cockpit and bodywork — and regardless of aesthetics, how every surface exists for a reason.

Modernising it therefore presents an interesting dilemma—change too little and you simply have a mildly modified 917—change too much and you lose the very characteristics that make it recognisable.

So where does a designer draw the line? …that became Tony's first challenge.

Process One: The Traditional Designer

For the first stage of the experiment, Tony tackled the Porsche 917 using traditional design methods — sketching and developing the design without any AI assistance.

Tony: There were a number of body variations of the 917, but it’s the 917K which I think is the version most people will think of. There’s a great line in a Bond film (you’d be amazed how much wisdom can be found in those films!) where the villain says the key to any good story isn’t who or when or what… but “why”, and it’s the reasoning behind the way the 917K looks which I find most fascinating.

That distinctive forward situated cabin was so they could push the driver as far forward as possible to fit the huge 4.5 and 5l flat 12 engines into the short wheelbase required to fit into Porsches existing architecture, the bubble-like wrap around windscreen was to give the forward positioned driver as much vision as possible whilst making the cockpit aerodynamic enough to deal with average 240Kmh speeds.

The big flat rear deck was to provide as much downforce as possible and cure the lifting tendency the earlier long tailed cars had at speed, and those little flip up spoilers at the end of the rear wings came about through someone looking at the way insects had splattered on the body whilst testing.

It’s a great example of function defining form. It’s not styled, it’s designed to do one job very well and that is to go very fast… and the slightly stubby, purposeful shape is a perfect example of how engineering can create something, if not beautiful, certainly incredibly striking.

Tony explored a number of iterations, allowing the proportions and character of the modern 917 to evolve through his sketches.

Tony: I tried not to look at pictures of a 917 when I started and instead rely on the distinctive features on the car that had stuck out in my memory. The forward positioned cabin and long rear deck always evoked in me a feeling of jet fighter for the road, so that’s where my thoughts began.

There would clearly be an amount of aesthetic input, I am a designer after all, but there had to at least appear to be some engineering reason behind the forms. It had to have the same downforce creating wedge shaped body and cut off tail, but I felt a more constructed aesthetic was in order.

I’ve been quite fascinated lately with the idea of having body panels that appear to float slightly off the structure to passively create pathways for air to flow through, using panel splits and gaps to form aero features, aerofoil sectioned surfaces that wrapped over the underbody and guide the air around giving the car a lithe, lightweight look, and these thoughts took form in the early doodles.

I wondered whether the graphic of the big front lights could be distilled down and recreated with LED light bars floating in an intake area, and that forward positioned cockpit with its wrap around screen simplified into one glass bubble surface.

Early on I also thought it might be interesting to use the panel separations to create a similar colour split to that in the gulf racing colours that the 917 is most known for, and that helped define the way the front wings wrap over the wheels and create a front spoiler affect between the lights.

The louvred venting on the top of the front wings is not unique to the 917 but always fascinated me, and I felt there was something interesting there that could be developed into a cool detail that (if set at an angle) would help enhance the “wedginess” of the body.

I thought the rear needed to be simple—open and uncluttered allowing the visual drama of the wedge shape to make its own statement and whilst exposing perhaps a little suspension and diffuser detail, not overdo it. I also liked the idea of a simple transparent acrylic flip up at the end of the wings that subtly echoes the aluminium spoiler added to the 917’s tail, but when illuminated would create the taillights.

Process Two: Designing with AI

Tony was now asked to start again with the same car and the same brief, but this time with AI as part of the design process.

Tony: When sketching, you are working towards refining an image or an idea you already have in your head, knowing roughly what you want and the end result desired, whereas using only AI and prompts can feel a little more like making the best of what you’re given.

I don’t think any designer starts off without a hint of preconceived direction, and my aim with the AI was to create a design with a similar feel to the one I’d sketched and see if the randomness of the tool would spark any new directions or ideas.

Using a series of prompts and iterations, Tony began exploring different interpretations of the modern 917.

AI Prompt: "Imagine a modern concept interpretation of a Porsche 917K, exaggerated wedge shape to body, small cabin with wrap around visor-like windscreen, minimalist soft forms, show car in gulf colours using the colour splits to define different panels".

AI Prompt: "Create a modern concept interpretation of a Porsche 917K, exaggerated wedge shape to body, small cabin with wrap around visor-like windscreen, minimalist soft forms, show car in Gulf colours using the colour splits to define different panels that create aero features".

AI Prompt: "Change windscreen and side glass to one wrap around screen".

AI Prompt: "Photorealistic image of car parked on Monaco Harbourside, orange parts to be orange carbon fibre, blue areas to be light metallic grey satin".

Tony: I began using AI to create 4 images from a prompt to get a spread of ideas, but initially what was produced was quite generic. I did admittedly have an idea in my mind of the design direction I felt I wanted the car to go so it was a little disappointing, but I quickly found by altering the prompts the results changed and I was able to tailor the output to produce images and ideas that I could combine, choosing the best bits, blending them, altering proportions, colours and components until I felt I was getting an interesting result.

The front quickly headed in a good direction with an interesting graphic, but I found it difficult to get to a rear end I liked from the stream originating from the original prompt. I was quickly able to create an alternative design though, with different prompts which gave a more dramatic style—and then was able to get AI to blend the rear of it with my primary direction.

In the process of imagining the different views, AI changed that rear subtly, but I quite liked the result, it was perhaps a little more “road car” than intended, but I kept it.

I liked the fact AI doesn’t produce exactly the image in my head, which led me to look in alternative directions and accept, and modify ideas that I may not have initially considered, but which I saw new potential in.

In a way it’s like a “ready, steady, cook” for design. Rather than combining set ingredients to get to a dish you know, you’re given random ingredients and it’s up to you to see what you can make of them.

The resulting vehicle from this route had a more product design look about it in terms of the form language, which I like, and I thought the treatment of the glass house that was arrived at was pretty interesting.

One of the benefits AI brings is the ability to quickly visualise the car in different scenarios and really get a feel for how its aesthetic might work in reality, and then very swiftly create views from different angles once that visual direction is established.

The Designer Becomes the Curator

This is perhaps the most interesting part of the experiment, because although AI may generate images, it doesn't remove the need for human critique. Tony still has to decide which results are interesting and which to discard. He has to recognise when an image has the right proportion but the wrong detail, or when an unexpected AI interpretation contains an idea worth pursuing. So, the role of the designer begins to shift. Instead of creating every element directly, the designer becomes a combination of creative director, editor, critic and curator.

Tony: The randomness of the results from prompts always surprises me, especially how the same prompt can produce wildly different results, and how a marginal change in wording can alter the output quite dramatically.

It can be both very inspiring and very frustrating at the same time. Because of this, it creates options that you might not think of. Whatever AI you use, it isn’t telepathic—it relies on the words you feed it, regardless of any vague ideas you might have in in the back of your head as result of your own accumulated knowledge.

I’d have previously found new design directions by wandering around design shops and looking at furniture and architecture, but it frees you from that sketching muscle memory.

Getting the prompt right is of course important, and I recall a few years back trying to get a CAS modeller to create a surface to match a component sketch I’d done—and try as he might it wasn’t going the right way. He finally turned to me and said, “You can explain it as much as you want, but you can’t make me understand!”.

Working with AI feels a bit like that, like trying to get a talented but rather unruly junior designer to follow direction and get to an idea that you can visualise clearly, that they see differently. It doesn’t so much replace a designer as promote them to a more managerial position where they are choosing and combining options to get to a result.

The “a bit of this, some of that, none of those, and all of the other” way that we all traditionally dreaded in sketch reviews, because it ruined the purity of our perfect (or so we thought) ideas, but in this case, we can use to our advantage.

The "Mixed" Input Approach: Combining Hand-drawn Sketches and AI-assisted Design

AI Prompt: "Concept Porsche Le Mans racing car with 917 influence".

AI Prompt: "Using front 3/4 view "Imagine side view of car".

Side view sketched over and proportions adjusted using Procreate.

A couple of side views to experiment with wheel colours, using previous side view with prompt: "Imagine concept Porsche 917 parked on Monaco harbour edge at dusk, make vehicle 10% longer, remove stickers, maintain wheel design but change colour to black with orange carbon fibre inserts".

Using the modified side view, I asked Vizcom to "Create ideas for rear 3/4 view of concept Porsche 917 road car, minimalist rear end and LED tail lights that wrap over the aerodynamic kick up and end of rear wing".

Not happy with any of the Vizcom results, I did a quick rear end doodle of where I wanted to go and fed that back into Vizcom with the same prompt. The result was questionable but had a few good points, so I sketched over that with Procreate.

The Procreate image was then added to the side view with the prompt "Combine side view of car in @1 with rear view @2 to create a new rear 3/4 view".

That image was then sketched onto with Procreate, then fed back into Vizcom with a prompt to "Imagine it on the Monaco harbour front, in grey, without graphics". The result was then sketched a bit further to get the transparent flip-up/light effect I was trying to achieve on the rear.

Returning to a front view, a bit of sketching over this image, added wheels from the rear view and then fed back into Vizcom.

Final Results: The Two 917s

Eventually, the experiment produced two interpretations of the same car. One was created through Tony Hunter's traditional design process, beginning with hand-drawn sketches, followed by a "mixed" input (hand-drawn/AI-assisted) approach, and the other AI-assisted process was driven entirely with AI prompts.

AI-assisted 917: Designed with Prompts

"Mixed" Input: Hand-drawn/AI-assisted 917

This isn't where we ask which one is better—instead, we asked Tony what he thought about the two results.

Tony: Both processes are enjoyable in their own way, though I would suggest for “dyed in the wool” designers who possess that little bit of creative OCD necessary for the job, the lack of control over AI can become a bit annoying!

AI can result in something either a bit derivative or utterly bonkers if left to its own devices, and it takes careful weeding and combining of its output to get a good result. It can of course be very amusing to see the strange and often silly variations that it can produce, and as a result it can quickly supply some very unexpected directions.

It’s also useful to see the design immediately in realistic surroundings, especially when you take it further and make a quick animation of the vehicle moving. Cars are not supposed to be static objects so the ability to see the shape moving and turning, and how light might cross the surfaces is incredibly useful.

One disadvantage is that it can give a false sense of credibility to a bad idea—skipping straight into apparent “shiny and parked at the side of the road” feasibility, without the thought development that would go into a sketch, where lines are positioned for a good reason.

Early in the design review process, the visual accuracy of AI images can also sometimes be a negative. A loose sketch that can be interpreted in different ways is often more likely to get picked for development than a perfect photographic image, which can be too specific.

I don’t think I’ll ever want to stop sketching, the creative process of exorcising stubborn ideas and thoughts onto paper or tablet is relaxing and meditative as well as productive. The love of that is part is why I wanted to get into the design business, and no matter how much more accurate designing with AI might get, the art involved in doing it by hand is always going to be more personal and fun.

My ideal workflow now would be the hybrid of the two routes, where an initial design sketch supplies a base direction, then gets explored, varied quickly and assessed using AI, with manual adjustments drawn by hand, then reintegrated into the theme with AI to create an easily understandable image to review. The benefits are that a more demonstrable thought development process can be seen (which is often asked for in sketch reviews, back to the “why” question).

It should always be remembered that getting to a visual direction is just the first stage in the design process. The 3D development, turning those images into feasible and ultimately marketable products in conjunction with surfacing, engineering, ergonomics, aerodynamics, CMF, manufacturing and all the other teams that come together to put a car on the road or track, is still something that needs a largely human input.

So, What Did We Learn?

There is an obvious temptation to turn an experiment like this into a contest, but that was never really the point. There is no winner.

The most revealing aspect of the experiment is that AI didn't replace the designer — it changed the designer's role.

Traditional automotive design is fundamentally an act of controlled creation. The designer decides what to draw and then draws it. AI introduces a degree of unpredictability. The designer asks for something, receives something (expected or unexpected), evaluates it, and decides whether the results contain an idea worth pursuing.

That can be frustrating… but it can also be creatively liberating. And perhaps that is where AI's most interesting role in automotive design currently lies — not as a replacement for the designer, but as a tool for generating “unexpected” possibilities.

"In a way it’s like a “ready, steady, cook” for design. Rather than combining set ingredients to get to a dish you know, you’re given random ingredients and it’s up to you to see what you can make of them."

The most interesting question now is… “what happens next?”

Tony: I think whilst you can produce a design solely using AI, it can be very useful to shake up thoughts and open new avenues of exploration. When it’s a more specific defined design target that you’re aiming for—when you want some real innovation—AI needs that human start point and guidance, so for me the ideal workflow is one where the AI tool becomes a partner in the design process (as used on the 3rd hybrid proposal).

Whether AI is used to refine and explore a set of pre-sketched ideas or simply throw out some alternative ideas to challenge preconceptions, and then be manually developed, it’s really useful for achieving design goals in a quicker more efficient way.

It doesn’t replace the designer though—it alters their role, supercharging the output, in much the same way that Photoshop initially did, followed by Alias—changing the way a designer worked. AI will similarly change the process going forwards.

One thing will remain though—in order to create great pieces of design, human input is always going to be desirable.

A Note of Appreciation from TorqArt

TorqArt would like to thank Tony Hunter for the time, creativity and considerable effort he has devoted to this project. Taking on the same design challenge through two different approaches required not only his considerable design experience, but also a willingness to experiment, explore new possibilities and share his process openly with our readers.

We're grateful to Tony for embracing the challenge and giving us a unique insight into the evolving relationship between automotive design and AI.

© 2025 - 2026 TorqArt. All rights reserved

Privacy Policy

C O N N E C T