Today we are talking about a surprising function – but to be used with caution – which are making those who experience artificial intelligence in design discuss: transform images into vector format.
Let’s find out how it is done (even with an example prompt), if the function is truly effective and for which activities to use it at this stage.
>> Do you want to stay updated and receive articles like this? Click here, it’s free
Image-to-cad with chatgpt: how much is it really useful?
In recent months, among those who experience artificial intelligence for the design, an interesting question has taken hold: Is it possible to transform an architectural image into a cad 2D design using chatgpt?
To try with method and constancy is Amir Hossein Nooriarchitect and digital popularizer who shared a series of surprising tests on LinkedIn. Use Chatgpt-4oAmir loaded architectural images similar to a prospective view and asked the model to generate one Bidimensional CAD representation of the same.
In response, Chatgpt generated a two -dimensional and vector image version of the image, which returns the essential geometry of the prospectus.
Image-to-cad with chatgpt: does it really work?
The short answer is: Depends. The system works interestingly on simple, clean images, with well -defined geometriesas a façade prospects in black and white or volumetric schemes with strong contrast. In these cases, Ai manages to recognize architectural elements, outlines and alignments, restoring a fairly coherent digital design.
But in the most complex cases – for example perspective views, stratified sections, plants with furnishings or retini – the limits current of the instrument:
- difficulties in recognizing real proportions and stairs,
- approximations in curved geometries,
- impossibility of managing overlaps or visual depths,
- need for a final manual control.
As Amir himself notes, We are not yet close to sufficient precision for direct use in constructive areasbut it is already possible to start using it as support for other phases.
Where can it be really useful?
- Quick translation of sketches for presentations: Do you have a conceptual view of a project and do you want to generate in a few seconds a vector stroke legible for a slide, a diagram or a graphic table? This process can accelerate the work.
- Conceptual display in the ideation phase: in the studio, it can be useful for converting preliminary sketches into digital forms to be manipulated or commenting together with the team.
- Obtain vector of architectural components: even if the result is not a complete technical drawing, chatgpt can be used for quickly extract individual objects – such as fixtures, furnishings, facade modules, profiles or decorations – from an image. It is an effective way for save time in manual recalculation and popular drawings and tables with vector elements ready for use.
Image-to-cad with chatgpt: example prompt
Here is a Example prompt to be used by loading the image that we want to give as a reference:“Create 2D flat elevation cad drawing of this view. There must be no sense of depth in the 2D cad drawing and no perspective lines are allowed.”
We see the results. Amir has documented his experiments with images before/after Directly on LinkedIn: I point out one here. The images show how, from a simple render or screen, the system returns a 2D “Flattened” prospectus. I advise you to take a look at the posts to evaluate with your eyes what works … and what not.
In summary:
- It is not yet a technical production tool.
- However it is A good ally to test, view and reason in digital form already starting from an image.
As always, the advice is: Don’t wait for it to be perfect to start trying it… Until next time!
Luciana
The weekly section “Architectural Prompting” is curated by the experts Luciana Mastrolia, Giovanna Panucci and Andrea Tinazzo
>> If you are interested in these issues, also subscribe to the Free LinkedIn Newsletter Ai & Design for Technicianslet’s talk about it here!
