Visualizing furniture with AI
Using the text-to-image generator Stable Diffusion, the carpentry workshop Dein Freund wants to create an image library of furniture designs as a basis for advising customers and designing furniture.
Responding quickly to a customer's request for a new piece of furniture is crucial. Customers want to know immediately if they are being served by an expert supplier. If not, they may choose another company that meets their requirements. Speed is therefore of the essence in the design process. Ideally, a visual representation should be created from the customer's description on the same day. However, time constraints are often a challenge for designers and often result in only a rudimentary sketch being produced. With the help of AI, the carpentry workshop Dein Freund aims to significantly speed up the furniture design process without compromising the creativity and originality of its designs.
Generative artificial intelligence has now entered the mainstream. However, its application is by no means trivial. That's why DAISEC is helping the carpentry workshop Dein Freund to enter the world of latent diffusion models, such as the text-to-image generator Stable Diffusion, and is accompanying them on their way to implementation. By using a design sketch in conjunction with a verbal description, also known as a 'prompt', the AI should be able to complete the sketch with colour and detail.
Diffusion models are a special form of artificial intelligence used to generate data. These models start with simple noise and modify it through a series of small steps until it becomes complex data such as images, text or music. A prominent and freely available example of such a model is Stable Diffusion. A major advantage of Stable Diffusion is that it requires less computing power because it does not work directly on the image, but first 'compresses' it and then processes it in the compressed space. Another advantage is the extensive ecosystem that surrounds Stable Diffusion. Thanks to numerous extensions, the tool's capabilities grow every day. One such extension is the ControlNet, which allows the otherwise random image generation process to be guided by a sketch. The newly created image retains the proportions, edges and corners of the sketch. This is particularly important in the design process. Finally, the graphical user interface 'Automatic1111' can be used to control Stable Diffusion in combination with ControlNet. DAISEC provided an introduction to Stable Diffusion for the carpentry workshop Dein Freund, carried out the local installation of the three tools mentioned and produced a best practice guide. As a result, the carpentry workshop Dein Freund is now able to present a product proposal to customers more efficiently and quickly.
For this use case, we have also published a news item. Read it here:
Dein Freund is a modern carpentry workshop based in Wedemark, near Hanover. Founded in 1999 by Wilhelm Freund, the award-winning company designs and builds exclusive furniture and undertakes interior and bathroom design. Part of the company's philosophy is to increase its own efficiency through digitalisation and to be able to respond to customer wishes in a targeted way.