Image & Text to 3D-printable models

During one of my Research & Design (O&O) classes, my students were tasked with creating a prototype of their model in 3D. To my surprise, they completed the task so quickly, it seemed almost unbelievable. When I asked how they did it, they showed me a website called MeshyAI, something I had never heard of before. Thanks to my resourceful students, I was introduced to this incredible tool!

They simply picked an image they liked from Google and, using MeshyAI, generated a fully printable 3D model in no time. This got me thinking: Are my students being smart by working this way? Why go through the painstaking, traditional process of modeling from scratch when there’s an easy and efficient alternative? By embracing tools like MeshyAI, they leave more time to develop other skills they find more interesting and relevant to their work. But is this the future of design? Let’s dive into how MeshyAI is reshaping 3D modeling and whether it’s the right approach for students and professionals alike.

Before diving in, I couldn’t resist giving it a try myself! While working on a Magic: The Gathering Angels Deck, I stumbled upon a blurry image of an angel on Google that caught my eye. Inspired by the deck, I thought, "Why not bring my Commander Card to life and print it to use as a visual token on the board?" So, I uploaded the image to MeshyAI, and here’s what I got: Image input: (click me)

MeshyAI output: (open Link to view the 3D-model in 3D)

This post is mainly to inform you that this tool exists. While creating detailed models yourself to your exact specifications may still be preferable, since AI can sometimes miss certain details, depth, or specific elements you want, MeshyAI is great for quick prototyping. It’s effective for generating general and even specific models based on text or image prompts, saving you time on the initial creation process. However, fine-tuning and modeling might still be needed for more precise results.

So let's start this post with the use of three AI's to test the 3D-moddeling output:

• Idea generation prompt with ChatGPT (you):

  "Create a fantasy medieval knight art description in text of a young adult Western male, with blonde, medium curly hair, blue eyes, and a strong athletic posture. He is fierce, looking strong and ready to go into battle, holding a grand claymore. He wears a black, tattered cloak. Make it 40 words long."

• > ChatGPT output: A young, athletic Western male knight with blonde, medium curly hair and piercing blue eyes stands fierce and ready for battle. He grips a grand claymore, his posture strong. A black, tattered cloak billows behind him, signaling countless battles fought.

  
  

Now that we have a text-based art description, let's use it to create a visual representation in Midjourney. We'll ask ChatGPT to generate 4 diverse variants of the description to explore different artistic interpretations and styles.

4 art prompts from ChatGPT (open/close me)

1: "A young, athletic Western male knight with blonde, medium curly hair and piercing blue eyes stands fierce and ready for battle. He grips a grand claymore, his posture strong. A black, tattered cloak billows behind him, signaling countless battles fought." 2: "A determined, athletic Western male knight with blonde, curly hair and piercing blue eyes stands tall and ready for war. His firm grip on a grand claymore and the torn edges of his black cloak reflect the weight of countless battles." 3: "A young, battle-hardened male knight from the West, with blonde, curly hair and intense blue eyes. He stands tall, holding his grand claymore, his athletic form radiating power. His black cloak, tattered and torn, marks the scars of a warrior's journey." 4: "A fierce, athletic Western knight with blonde, medium curly hair and sharp blue eyes. His muscular frame is set for battle, a grand claymore gripped tightly in his hands. His tattered black cloak whips in the wind, a testament to his warrior's resolve."

> Midjourney Output (click me):

I find the first prompt we generated that gives the best results, but it's not a full-body pose. Let's make that clear in the prompt: "please add a white background and full body pose"

> ChatGPT Output:

A young, athletic Western male knight with blonde, medium curly hair and piercing blue eyes stands in a full-body pose, fierce and ready for battle. He grips a grand claymore with both hands, his strong posture radiating power. A black, tattered cloak billows behind him, signaling countless battles fought. The background is pure white, emphasizing the knight’s heroic figure and unwavering stance. > Midjourney Output:

We could dive deeper into art styles, prompt adjustments, and other details, but that's not the focus of this blog post. Instead, let's generate some models using MeshyAI. I will upload the first image and provide the corresponding text prompt to see the different results in terms of quality and check if it generates the feet correctly!

>Text to 3D-output:

>Image to 3D-output:

As we can clearly see, this picture to 3D-model is horrible. But there is a good conclusion to make out of this. Those are all black collors and creating depth and detail is very hard that way. Imagen writing software that transforms a 2D-pictures into 3D models. For that, you can ask ChatGPT how that exactly works, it's insane! But! To make this better, for the sake of the blog. Let's redefine the prompt and ask ChatGPT: "Make this prompt (our character prompt) super clear, high black and white depth details" >ChatGPT Output:

A young, athletic Western male knight with blonde, medium curly hair and piercing blue eyes stands in a full-body pose, fierce and ready for battle. He grips a grand claymore with both hands, exuding strength and determination. His posture is powerful and commanding. A black, tattered cloak billows behind him, representing the wear of countless battles fought. The image features sharp contrasts with deep blacks and bright whites, adding depth and intensity. The pure white background highlights the knight’s heroic figure, accentuating his unwavering stance and presence.

>Midjourney Output:

It’s not perfect, but let’s see how it turns out. >MeshyAI output:

Much better! It’s not quite perfect yet, but it’s definitely coming together. Imagine crafting it by hand! You could even generate a rigged body and animate your character, but for that, a more natural pose would be needed. Let’s wrap up the blog post!

❗The models I created were essentially completed in under 2 minutes (idea generation to 3D-model)! However, more time dedicated to refining the models/prompts would lead to significantly better results.❗

What you learned in this post:

• Quick Prototyping with AI: MeshyAI enables fast 3D model creation, ideal for quickly visualizing ideas in early design stages.

• Enhanced Results with Combined Tools: Using ChatGPT for prompts and Midjourney for visual references helps create detailed descriptions for MeshyAI to generate more accurate models.

• Speed vs. Detail: AI tools like MeshyAI offer speed but may require refinement for fine details, making them excellent for prototypes.

• Rigging and Animation Potential: MeshyAI can generate rigged and animated models with proper input.

Conclusion: In this post, we've explored how AI tools like MeshyAI, ChatGPT, and Midjourney can reshape the design process. From quick 3D-prototyping with MeshyAI to generating detailed visual prompts with ChatGPT and Midjourney, these tools allow students and professionals to rapidly visualize and refine their designs, freeing up time to focus on more creative and technical aspects.

While MeshyAI excels at fast prototyping, the fine details may require some adjustments. The combination of AI tools opens up a new realm of possibilities in design education, making it easier to experiment, iterate, and bring ideas to life.

If you found this interesting or have any questions,

don’t hesitate to reach out! I’d love to help, contact me today!