3D Modelling & Artificial Intelligence - creating aesthetically pleasing 3D models | Articles

As someone always eager to learn and experiment with new aspects of 3D printing and modelling, I recently became intrigued by the surge of AI-labelled, visually stunning models shared by other creators. The blend of artificial intelligence with 3D modelling has sparked my curiosity: how are these impressive creations made? This article documents my ongoing journey from complete novice to someone capable of using AI tools to design aesthetically pleasing and functional 3D models. I’ll be updating this narrative as I explore further, so check back for new insights. If you have comments or feedback, I’d love to hear from you!

Background

I am not an artist and have virtually no sculpting skills, but I aspire to develop 3D models that are both beautiful and practical. AI, as I’ve come to realize, can do much of the heavy lifting in the sculpting process. My role is to provide core inputs and requirements, then work alongside the AI to draft the results and then I can touchup/modify to complete the refined model. What I’ve discovered so far is inspiring and has me excited about the potential of these new tools.

Where I started

My initial AI test project was to create an aesthetically pleasing rain gauge that was functional. The rain gauge is to incorporate natural looking elements (flowering plant and a humming bird) plus an actual 21cm by 4.4cm diameter glass rain gauge. The rain gauge is to be placed within a planter so the plant model needs to be spiked on the end.

The process I followed started with the text description of what I wanted to create and then using that as prompts move onto having AI create 2D images. I would then utilize 3D modelling AI to turn the 2D images into 3D models with some tweaking and then finally import into CAD software (Fusion 360) and do the final modifications (scaling, functional changes and prep to import into slicer software.

Text Prompts

The creation of the requirements in a text format if completed correctly will provide the AI tools with the majority of information to efficiently create the initial draft images that are more closely matching what you are visualizing in your brain.

There are some key items to be included in the text prompts.

background - keep it plain
lighting- soft lighting
resolution - high resolution
perspective - show depth of image to provide the 3D perspective
description - concise description that avoids distractions (eg logos)

Example prompts I used

Humming Bird Text Prompt – “Generate a high resolution image of a realistic hummingbird in flight with grey coloration on a white background that is symmetrical that can be used to help AI generate a 3D model.”
Tulip Plant text prompt – “Create a high resolution image of a tulip like plant that has a main vertical stem with a spiked bottom end. From the main stem have multiple tulip flowers on different planes around the central bloom at varying heights and stages of growth. Have the image optimized for AI to use for generating a 3D model.”
Template version of prompt - "Create a realistic high resolution image on a plain white background with soft lighting of "description goes here" that shows a 3/4 perspective that will help with the creation of a 3D model."

2D Image Generation

Humming bird 2D image generation - I used Microsoft Copilot to generate a realistic, detailed 2D image of a white hummingbird (with no background) using text prompts. Copilot produced a high-quality image, which I downloaded in PNG format.

Multi headed tulip plant image generation - I also utilized ChatGPT to create a 2nd 2D image of a tulip like plant with multiple flowers (tulips do not do this in nature).

3D Model Generation

With the two AI generated images (humming bird and plant) I explored a few different browser-based AI modelling tools to generate 3D models:

Meshy.AI: This web application is beginner-friendly, with helpful tutorials. The free subscription lets you experiment, though downloading 3D models requires a paid plan. I used my hummingbird image as an “image to 3D model” prompt, and Meshy.AI swiftly generated a 3D model that could be rotated and edited in-browser.
Rodin AI: Another browser-based tool, though access can be inconsistent. After signing up for a free account, I uploaded my hummingbird image and used the basic generation and material tools. With the free tier, I could download the resulting model in STL and PNG formats. While it lacks advanced features without a subscription, Rodin AI did a solid job, allowing me to review the model for inconsistencies.
Printpal.IO: This tool offers a clean, user-friendly interface and allows free downloads of basic 3D models in formats like OBJ, GLB, PLY, and STL. The models generated are quick and functional, though more detailed refinement requires a paid upgrade.

The outputs were fairly similar but the AI 3D models I used for the final steps was based on which application allowed me to do a free download.

3D Model Updates and Touchup

I used Fusion 360 as the main integration and CAD tool for additional functional changes. I know there are other tools but this one I am familiar with.

Integration - I imported the two (plant and hummingbird) 3D generated models into Fusion 360 as meshes. In Fusion 360 I placed the humming bird near one of the tulip flower heads with beak inserted and then combined the two models into a single mesh.
Scaling – I scaled up the model so that it was now at the appropriate size for any additional functional changes.
Functional Changes – I modified the main tulip flower to allow the glass rain gauge to be inserted and held securely while allowing visibility to the entire length of the glass container.

Tools (used so far)

Copilot - Microsoft AI tool that is useful for creating a 2D image as well as helping to create the prompts. I also used this AI to help draft the content and imagery for this article.
Chat GPT - AI tool that has a lot of different capabilities that I used for 2D image creation
Meshy.AI - online web browser tool (has a free subscription - does not allow download of 3d models unless you upgrade to paid version) that is an easy starting point for using AI to create 3D models.
Rodin AI - online web browser based 3d modelling application. The free subscription - does allow for at least one of 3d models (either as OBJ or STL files plus PNG format. A number of refinement functions are only unlocked with a paid subscription.
Printpal.IO - online web browser based 3d modelling application. The free subscription - does allow for 3d models (either as OBJ, GLB, PLY or STL files format) to be downloaded. A number of refinement functions are only unlocked with a paid subscription.
Fusion 360 - CAD tool for which I have a hobbyist license so not all functions are available

Reflections and What's Next

My journey so far has shown that AI can empower creators like myself who have minimal artistic or sculpting backgrounds yet can produce models that are both beautiful and functional. The integration of AI into 3D modelling unlocks new possibilities and lowers the barriers for entry into this creative field. I am continuing to experiment with different tools and workflows and will share updates as I learn more.

While Fusion 360 was great for the functional (specific measured item change) I suspect there are better sculpting tools (meshmixer, mudbox, blender) that may be better suited to the artistic changes.

If you’re also exploring AI in 3D modelling, I encourage you to reach out with your insights or resources. Let’s learn together!

3D Modelling & Artificial Intelligence - creating aesthetically pleasing 3D models