Crumly

Crumly is an app to generate instant 3D visuals for design ideas in seconds.
It makes it very smooth and easy to generate quick, clear visuals from any 3D model.

It works in 3 simple steps:

  • Upload your 3D model.
    Upload your model

  • Set the view you want to capture.
    Set the view

  • Write a simple prompt and get your result.
    Generate the visual

You can submit one of the several commonly used 3D file-formats. There are inbuilt tools to add missing details and assist you in generating desired result.
You will be surprised at the quality of results you can get even with a crude block model.

Give it a try at:
crumly.vercel.app

Inspiration

I have worked with several architects and other professionals who design spaces. They are often very creative people who express frustration at the clunkiness of design tools at the early stage of their design process. It becomes an encumbrance in their creative process. Besides that it is also very expensive and the ones who do put in the budget and effort are left with redundant work if the project scope changes.

What it does

It is a tool to help communicate spatial/physical design ideas quickly. Architects and designers often work within the context of an established physical space. This high level arrangement is then tweaked, refined and reworked by the designers in their design/execution process. eg:

  • Architects may work within the context of an existing complex to create new buildings.
  • Interior designs have little agency over the structural members of the space they are working in.
  • Set designers have physical limitations dictated by camera/lens and how the shots are framed.

Beyond these constraints there is a wide range of possibilities that the designers and ideate in to come up with their unique solutions. eg:

  • Architects experiment with different layouts to see how it responds to the context.
  • Interior designers will explore different themes to impart character to the spaces they design.
  • Set designers are often tasked with delivering spaces which will act as a backdrop to a scene. The position, orientation and theme are defined with respect to the camera, actors and the plot.

Communicating these ideas become challenging when they don't have a good format to overlay their ideas over the existing physical setup.

Current ways of doing that involves expensive, time-consuming and laborious process of capturing the space in a digital format, documenting every design idea, communicating it with 3D modelling professionals and extracting presentation materials from the generated models. The models generated need to be of a good quality to be able to extract effective communication/presentation materials from it.

It is often the only way the "big idea" on the design is shared with everyone on the team. At the early stage however developing the high fidelity models and presentations take a lot of time and effort. Moreover the high-quality materials will go to waste if some other upstream aspect of the project changes.

How we built it

It's built as a web app made with nodejs + React running on vercel. There is a 3D model viewed embedded on the page using threejs which allows the user to interact with the model and set the view. They can capture the view and which is loaded to the React canvas to allow making annotations on the app. The edited image is then uploaded to server which interfaces with a stable diffusion model along with text prompt to generate a result image. Convex is used for authentication (using Clerk), running the backend service, database and file storage.

Sample Model Files

What's next for Crumly

This is just a start. We are generating limited resolution images for now. We will be exploring this in discussion with working professionals to understand how it fits into their workflow.

Built With

Share this project:

Updates