Meshy — Is It The Ultimate 3D Model Generation Tool to Turn Simple Text or 2D Images into 3D Game Models?
Generative AI is an emerging field that has not just redefined our concepts of AI but has also pushed the boundaries of creative innovation. Meshy is one such tool that is leveraging the generative AI potential to provide an innovative way to create 3D models (game assets) from simple text, 2D images, or untextured models. It claims to revolutionize the 3D content creation ecosystem. Therefore, this blog is all about Meshy, exploring its capabilities with a few experiments and evaluating its claims realistically. So, let’s get started!
Meshy — A Quick Overview
Meshy is an online generative AI toolbox that provides an intuitive and effortless way to create 3D game assets. With generative AI, Meshy is able to turn text-based descriptions into 3D models, 2D images into 3D images, and untextured models into 3D textures.
Meshy distinguishes itself by providing an autonomous and instant platform to create high-quality 3D models within minutes. All you have to do is provide the basic instructions and leave the remaining matter to it. It will craft the required 3D model as per your input.
Let’s now bring Meshy into the evaluation round and test its claims with some real tests.
Experiment #1. Testing Meshy Text to 3D Model Capability
Meshy claims to turn a text description into a high-quality 3D model. So, this experiment will evaluate this aspect of Meshy.
I asked Meshy to design “a table with a laptop and a cup of coffee” and set the style as “ultra-detailed”. After execution of the request, it provided the below 3D model.
Looking at the model, the laptop was actually a PC monitor, and the cup of coffee did not seem to be from this planet. Besides that, the graphics were all blurry and gamified. So, I don’t think it reflected the model as per my text correctly. So, I customized my request a bit to bring more clarity by asking it to avoid low-quality and blurry aspects. The response I got was even more bizarre than before.
The laptop was in a cut shape, the cup of coffee got smashed, and the table got even more dirty. I think it tried to make the model as gamified as possible. Overall, I was not really impressed by its text-to-3D model output, but maybe it looks good to game developers who like these blur shapes.
Experiment #2. Testing Meshy 2D Image to 3D Model Capability
After testing text to 3D model, the next thing that is a highlight of Meshy is the 2D image to 3D model. However, the Meshy.ai webpage was not directly allowing access to this feature. I had to join their Discord server to use the feature.
After joining their discord server and watching their tutorial on how to use this feature, I tested its ability to convert a 2D image I got from Behance.
After waiting for around 10 minutes, I finally got the 3D model of the image.
I would say that it did turn the image into a 3D model with maximum emphasis on ensuring proper attention to details. I cannot call it a very high-quality work, but it is still impressive that we can create 3D models from images within minutes.
Experiment #3. Testing Meshy AI Texturing Feature
Meshy also provides a dedicated AI texturing feature that claims to generate high-quality textures for 3D models using text prompts and concept arts. So, my last experiment was to evaluate this feature.
I used the .gib file from my above 3D model and began a new AI texturing project.
In the text prompt, I asked it to design “a evil wonder woman with red hair”. The output model made me amazed and shocked.
Meshy did put the outfit of wonder woman similar superhero, but I don’t know whose face it inserted on the avatar. I was hoping for a more carton-based face. But still, it’s an impressive work. Just by providing the right commands, game developers or graphic designers can surely design highly graphical 3D models.
Final Words — Meshy is a Companion for Game Graphic Designers
After conducting three experiences on Meshy, I can say that it’s an innovative generative AI toolset for 3D game assets. It can design great quality 3D models through text, 2D images, and concept art. However, its features are still in development, and there is a learning curve to master its usability. So, game graphic designers should definitely give it a try and see how well it fulfills their needs.
Ranking:
- Accuracy: 3/5 — Acceptable 3D models.
- Ease of Use: 3/5 — Learning curve and the need to input the right commands for the required output.
- Customization: 1/5 — No API to integrate it into your application.
- Scalability: 4/5 — Can handle any complex design requirements.
