A new way of collaborating with AI
The creation and refinement of evocative and brand-aligned product geometry is a coveted craft. As both the volume and velocity of inspiration and insight expand, the challenge of creating novel and relevant designs becomes more complicated as industrial designers must filter more and more information. Like any artist, the industrial designer is tasked with fusing inspiration and new insight into an object of desire. At PCH Innovations, we’ve seen clients struggle with this challenge time and time again. Design teams are increasingly inundated with data and under extreme pressure to develop appealing 3D concepts that win hearts and minds.
We believe that AI-infused design tools can assist industrial designers by abstracting the complexity of design-relevant information and by using generative models to bridge the gap between 2D ideation and 3D design. We see a future where all product creators are empowered by AI to express innovative ideas. Where AI-assistants will enable fundamental shifts to product design processes by removing barriers to creative expression.
That's why we're excited to contribute to this new reality with BLANK — a cloud-based generative design tool that removes complexity from surface and proportion generation tasks.
BLANK enables product designers to:
- Easily generate and edit proportional 3D models
- Customize their development process with semantic control
- Incorporate their brands' 'design DNA'
- Dynamically navigate the data in their design portfolio
Easily generate and edit 3D models
Depending on the approach, traditional 3D concept design tools can require multiple hours to create a product concept form. These tools require users to have the expert technical skills and aesthetic understanding to make the creative leap from an idea to a formed 3D concept model. For professional industrial designers under pressure to create multiple concepts designs quickly, this is often a significant creative and technical challenge.
To address this challenge, BLANK's generative AI algorithm is trained on historical 3D designs and their associated meta-data to align the system with the context of the users’ targeted design space.The algorithm works by encoding 3D designs and metadata into a custom generative adversarial neural network (GAN) architecture, which we call XDGAN.
GANs are a class of deep-learning architecture that can generate new data instances to resemble the data they are trained on (read a more detailed explanation in AI for Designers). In the case of BLANK, each time a trained XDGAN model is inferenced, it generates a 3D product concept.
After training, BLANK enables designers to generate and edit unique, and contextually relevant, 3D concepts in milliseconds. BLANK’s generative algorithm learns features in the training data that are used as semantic parametric controls for manipulating designs. This enables any non-skilled user to quickly generate, edit, and explore potential 3D design concepts in real-time without having to edit complex surface sketches or manipulate BReps.
BLANK enables any non-skilled user to quickly generate, edit, and explore 3D design concepts in real-time without having to edit complex surface sketches or manipulate BReps surfaces.
Customize your semantic control
Winning product aesthetics are hard to create and extremely valuable. Consumers consistently rank aesthetics among the three most important attributes in product choice. In most industries, the process of aesthetic product design and testing requires designers and strategists to manually and visually process increasing volumes of historical product design, architectural logics and market data. This is both expensive and time-consuming. Furthermore, high-stakes design decisions are often made by subjectively decision-makers rather than through objective data-driven processes.
For example, in the automotive industry, the costs for designers manually developing and screening hundreds of early-stage aesthetic vehicle concepts can exceed $100M for a single vehicle manufacturer. To get feedback on the aesthetics of automotive designs, design teams ask consumers to evaluate design variations using in-person A/B testing. These customer clinics involve physically bringing target consumers to a central location to evaluate designs and rate them using well-established benchmarks such as semantic scales for sportiness, appeal, innovativeness, and luxuriousness. This is a significant source of waste in the design process. The application of data-driven design assistants will enable dramatic process improvements in this field.
BLANK enables designers to manipulate 3D concepts using semantic design controls related to product perception. This enables design firms, OEMs, and creators alike, to generate new forms that incorporate feedback from the start of any new product development process.
By training a BLANK generative model using 3D design data with metadata related to product performance, regional market research, sales figures, or qualitative consumer feedback, designers can develop concepts that leverage multiple dimensions of feedback into their selected concepts. For this reason, the semantically driven controls are learned by BLANK's AI using both supervised and unsupervised methods. That is, a design team can choose to train the system with metadata relevant to specific characteristics (e.g. market/segment performance, consumer perception), and, conversely, controls can be learned from unlabelled patterns in the data. This creates multiple options for design teams who want to customize their concept development workflows with the use of AI systems.
The semantic controls allow blending between two 3D models e.g. between a BMW (x) and an Audi (y); tuning by label properties e.g.-30% Sedan, +20% Coupe or -40% BMW, +15% Audi; or tuning by emergent properties e.g. by local product features, such as cabin or wheel size.
Incorporate your brand's 'design DNA’
Industrial designers spend a significant amount of time considering the aesthetic response to their idea. Designers must generate concepts of form that must both appeal to the future needs of their customers and align with the progression of their employer's design DNA and brand. Being able to manipulate the aesthetics and attributes of form to achieve a particular impression or feeling in a consumer is a critical skill for designers. The ability to communicate emotion through form and pair it with proportional attributes is the ability to control the design language of objects.
Leading brands spend significant resources developing product designs that communicate target ideas to targeted consumers. Will customers be prompted to buy? Will they associate the aesthetic and functional stimulus of the design with the characteristics they see in the brand? These questions make or break the commercial success of many products. BLANK's novel data interaction enables designers to infuse the design DNA of their brand (or someone else's) directly into a 3D concept in milliseconds. Designers can easily and quickly generate designs that are more aligned with a specific brand or market eg. more "Cadillac-like" or “Korean-market-like” and can choose to manipulate the design by adjusting bespoke controls related to semantic definitions, making the design "less aggressive”, “sexier”, “younger”, or “more mature”.
By leveraging our unique generative model that learns to represent a given corpus of 3D training data, we enable a novel understanding and communication of a brand’s 3D design language or their ‘design DNA’. This is something that currently resides in the subjective, tacit knowledge of a company's designers and is thus difficult to incorporate in workflows at scale.
With BLANK's unique capability, designers can incorporate quantitative and qualitative market data directly into initial product concepts and align with their brand’s aesthetic design language.
BLANK's novel data interaction enables designers to infuse the design DNA of their brand (or someone else's) directly into a 3D concept in milliseconds.
A tool for design, not just designers
We are building a tool to augment innovation by dissolving the human-machine interface and giving product creators the power to discover design opportunities hidden in data. We see a future where all product creators are empowered by AI to express innovative ideas and share them with their community. We aim to empower all innovators, from global automotive OEMs to backyard inventors, with the ability to visually articulate their 3D design ideas and share them with the world.
Interested to learn more? Reach out for a product demo today!