AI Can 3D Print Whatever You Ask For!

Summary

The video explores the journey of using artificial intelligence to generate 3D models from text descriptions. After testing various AI tools, the creator finds 'Mirage' to be the most effective, producing decent meshes without the need for downloads. The video demonstrates how to refine these models in Blender for 3D printing, highlighting the potential of AI in simplifying the modeling process. The host expresses excitement for the future of AI in 3D modeling and invites viewers to share their experiences.

Takeaways

• 🚀 Rapid advancements in AI are enabling new capabilities, such as AI-generated 3D models, which have not been seen before.
• 🎨 AI can now be used to create various forms of art, including books, images, and music, but 3D modeling with AI is a recent development.
• 🔍 The speaker dedicated weeks to find a way to generate 3D models from text descriptions, aiming to 3D print an AI-designed object.
• 📚 Initially, using chat GPT to write STL code for 3D models was explored, but it was limited to creating cubes or cuboids.
• 🍏 Another project, Point e, was able to generate more detailed shapes but was limited to certain objects like apples or donuts.
• 🌟 The Mirage AI tool was found to be the most effective for text-to-3D model conversion, producing decent meshes without the need for downloads.
• ⏱ Mirage generates models in about 3 minutes, which is longer than some other methods but yields better results.
• 🔗 Adding a reference photo to the text input in Mirage can significantly improve the accuracy and quality of the generated model.
• 🛠 The script provides a tutorial on how to clean up AI-generated models using Blender before 3D printing them.
• 🔄 The process involves removing separate parts, remeshing, and possibly filling gaps before applying modifiers for a smoother finish.
• 📦 The final cleaned model is exported as an STL file, ready for slicing and 3D printing.
• 🌐 The script concludes by emphasizing the potential of AI in 3D modeling and the speaker's excitement for future developments in this field.

Q & A

What was the primary goal of the speaker in the video?

The primary goal of the speaker was to 3D print something designed completely by an AI.

Which AI tool did the speaker find most effective for generating 3D models?

The speaker found Mirage to be the most effective AI tool for generating 3D models from text input.

What challenges did the speaker face when using ChatGPT to generate 3D models?

The speaker faced challenges with ChatGPT as it could only generate cubes or cuboids, regardless of the requested model.

Why did the speaker consider the project Point E, and what was the outcome?

The speaker considered the project Point E because it could generate a cloud of points to create more detailed shapes, but it only produced good results for simple models like an apple or a donut.

How does Mirage improve the quality of the generated 3D models?

Mirage improves the quality of the generated 3D models by allowing users to add reference photos, which drastically enhances the output quality.

What steps did the speaker recommend for cleaning and preparing the 3D models for printing?

The speaker recommended using Blender to clean and prepare the 3D models, including importing the model, deleting separated parts, using remesh and subdivision surface modifiers, and filling gaps with basic shapes.

How does adding a reference photo impact the 3D model generation in Mirage?

Adding a reference photo provides a clearer context for the AI, resulting in more accurate and detailed 3D models.

What limitations did the speaker find with the current AI-generated 3D models?

The speaker found that some models were too complicated for the AI to generate accurately, and the quality could be inconsistent.

What are some potential applications of AI-generated 3D models mentioned in the video?

Potential applications include enhancing 3D modeling routines by providing basic shapes quickly, creating concept art, and producing simple models for immediate printing.

What future developments in AI and 3D modeling does the speaker anticipate?

The speaker anticipates continued improvements in AI technology, leading to better and more detailed 3D model generation over time.

Outlines

00:00

🚀 AI-Driven 3D Modeling Breakthrough

The script introduces the rapid development of artificial intelligence and its emerging capabilities in 3D modeling. The author recounts their quest to 3D print an AI-designed object, highlighting the challenges faced in finding a suitable AI tool. They experimented with various AI systems, including chat GPT and Point e, but encountered limitations in generating diverse 3D models beyond simple shapes. The breakthrough came with the discovery of 'The Mirage,' an AI that converts text descriptions into 3D models with remarkable accuracy and detail. The author emphasizes the ease of use and the high-quality results produced by Mirage, which also allows for the inclusion of reference photos to enhance the modeling process.

05:01

🛠️ Refining AI-Generated 3D Models for Printing

This paragraph delves into the process of refining AI-generated 3D models for 3D printing using Blender, a free and user-friendly software. The author guides viewers through importing the AI-generated model, repairing the mesh, and using various modifiers to enhance the model's detail and smoothness. They also discuss the importance of using clean reference photos to improve the AI's understanding and the model's accuracy. The script concludes with instructions on how to export the model as an STL file for 3D printing, emphasizing the potential of AI-generated models as a valuable addition to the 3D modeling workflow. The author expresses excitement for the future of AI in 3D modeling and invites viewers to share their experiences and thoughts on the topic.

Keywords

💡Artificial Intelligence (AI)

Artificial Intelligence refers to the simulation of human intelligence in machines that are programmed to think like humans and mimic their actions. In the context of the video, AI is central to the theme as it discusses the rapid development and application of AI in creating 3D models from text inputs, showcasing how AI is advancing in areas traditionally dominated by human creativity and technical skill.

💡3D Modeling

3D Modeling is the process of creating a mathematical representation of any surface of a three-dimensional object via specialized software. The video focuses on the evolution of 3D modeling with AI, where the script mentions the lack of major breakthroughs in this area until recent developments, highlighting the potential of AI to revolutionize the field.

💡3D Printing

3D Printing is an additive manufacturing technique where a three-dimensional object is created by laying down successive layers of material. The script discusses the goal of 3D printing an object designed entirely by AI, emphasizing the practical application of AI-generated 3D models and the potential for streamlined creation and production processes.

💡Text-to-3D AI

Text-to-3D AI refers to AI systems that can interpret textual descriptions and generate three-dimensional models accordingly. The video script describes the journey of finding a suitable AI tool that can convert text descriptions into 3D models, with 'The Mirage' being highlighted as a successful example of such technology.

💡STL Code

STL Code stands for Stereolithography, which is a file format native to the stereolithography CAD software created by 3D Systems. In the script, it is mentioned as the initial method attempted to generate 3D models through AI, indicating the technical aspect of how AI can interact with 3D modeling software.

💡Point Cloud

A Point Cloud is a collection of data points in a coordinate system, often used to represent the external surface of an object in 3D space. The script refers to a project called 'Point e' that uses point clouds to generate detailed shapes, illustrating an alternative approach to creating 3D models with AI.

💡The Mirage

The Mirage is an AI tool mentioned in the script that creates 3D assets based on text input. It is highlighted as the most successful tool the narrator found for generating decent 3D meshes from text, emphasizing its role in the advancement of AI in 3D modeling.

💡GLB File Format

GLB is a binary file format for 3D models that is a subset of the GL Transmission Format. The script mentions downloading 3D models in GLB format, which is a practical step in the workflow of using AI-generated models for 3D printing.

💡Blender

Blender is a free and open-source 3D computer graphics software toolset used for creating animated films, visual effects, art, 3D printed models, motion graphics, and computer games. In the script, it is the software used to clean and prepare AI-generated 3D models for 3D printing, showcasing its utility in the 3D modeling process.

💡Remesh Modifier

In the context of 3D modeling, a Remesh modifier is used to create a new mesh from selected vertices, edges, or faces. The script describes using the remesh modifier in Blender to fix issues with the AI-generated 3D model, demonstrating a common technique for refining 3D models.

💡Subdivision Surface Modifier

Subdivision Surface is a modifier in 3D modeling software that smooths a mesh by adding more geometry. The script mentions using this modifier in Blender to smoothen the AI-generated model, which is a common practice for achieving a more refined and polished appearance on 3D models.

Highlights

Exploration of AI in 3D modeling, a field that has not seen major breakthroughs.

AI's current capabilities in creative tasks such as writing books, creating images, and composing music.

The challenge of generating 3D models from text, a task that has proven difficult.

The introduction of a solution that allows AI to design 3D models from text, successfully 3D printed.

Teaching viewers how to generate and 3D print models designed by AI.

Review of Andrew Singh's video using chat GPT to write STL code for 3D models.

Initial success with generating a cube using chat GPT, followed by failure with more complex shapes.

Evaluation of the Point e project for generating point clouds for 3D shapes.

Disappointment with Point e's inability to generate models beyond basic shapes like apples or donuts.

Discovery of The Mirage, an AI tool that converts text to 3D models effectively.

The Mirage's user-friendly interface and free account setup.

The Mirage's ability to generate detailed 3D models from text input, often surpassing previous methods.

The option to include a reference photo in The Mirage to improve model accuracy.

The recommendation to use clean, background-free photos for better AI understanding.

The occasional unpredictability of AI models, even from well-known platforms.

The Mirage's additional feature of creating concept art quickly.

A step-by-step guide on cleaning and preparing 3D models for printing using Blender.

Tips for improving model detail and filling gaps before adding a remesh modifier.

The use of subdivision surface modifier to smoothen the model before exporting as STL.

The excitement for the future of AI in 3D modeling and the anticipation of improved results.

Invitation for viewers to share their experiences with text to 3D AI and suggestions for video content.

Transcripts

artificial intelligence is developing

rapidly but never before was it able to

do this

but let's move back to the beginning

playing with AI has been available for a

while you can ask it to write a book

create an image or compose music for you

but in terms of 3D modeling we still

haven't experienced any major

breakthrough

this however is likely to change soon

I dedicated the last few weeks to the

new goal I wanted to 3D print something

designed completely by an AI

in order to do that I first had to find

a way to generate decent 3D models from

nothing but text

this was not an easy task but in the end

I came up with a solution that actually

worked the way I wanted

finally the time has come to share both

my struggles and the final outcome

I will also teach you how to generate

and 3D print models like this one so

keep watching

finding an optimal technique was quite

an issue

I came across the video from Andrew

Singh where he uses the chat GPT to

write an STL code for a 3D model

this looked very promising and the outer

demonstrated this technique on a cube as

a proof of concept

I opened chat GPT and also asked it for

a cube

[Music]

and it worked or so I thought

I was very excited so I tried to ask for

something else a simple flower pot for

example

but after running the code well you can

probably guess what's coming next I was

unable to get any other result than a

cube or a cuboid no matter what I asked

for

unfortunately this method was not what I

was looking for nice try though

the other project I found called Point e

looked quite impressive it is able to

generate a cloud of points creating a

more detailed shapes

again my enthusiasm skyrocketed but only

for a second after I tried to receive

anything else than an apple or a donut

I really tried to go easy on it asking

for basic models like a pier or a

traffic cone

but the outcome was terrible

I gave it one more chance this time

choosing the recommended search I

thought to myself okay since it's

recommended at least that should work

right

[Music]

I tried with few other available AIS and

honestly I almost gave up

but then I came across

The Mirage

Mirage is the closest text to 3D model

AI that I've seen so far

it creates 3D assets based on the text

input and it is the only tool I found

that actually comes up with a decent

mesh

you don't need to download anything and

starting a new account is free check the

description below for a link

after you open the app go to apps

section on the left

now choose text to 3D asset

to get a model you can simply write down

what you need and optionally add a

picture

it takes about 3 minutes to generate

which is a bit longer than previous

methods but damn the outcome is best

I've seen so far I mean compare this

flower pot with their previous ones

no comments

on top of that Mirage allows you to

easily download the 3D files which is

not possible with few other AIS out

there

[Music]

just click on your asset expand the

download section and choose glb

I will show you what to do with it

further in this video

I've been experimenting with few models

and the quality is actually impressive

[Music]

I noticed that adding a photo for a

reference can drastically improve the

results

here is a model generated with text only

and here is the one that included a

reference photo on top of the text

command

also I definitely recommend using clean

photos with no background as they come

out way better

like you see here when I loaded a

picture of a cat with a strong

background the AI could not properly

understand what I was looking for

not every request I made was instantly

successful some models were just too

complicated but to be fair that's fully

understandable even the big and

well-known AIS like mid-journey can be a

bit unpredictable sometimes as this

technology is still being developed

on top of that Mirage also allows you to

create sick concept art in matter of

seconds

in this video I'm Only focusing on the

text to 3D feature however I seriously

recommend you to have some fun with

other sections of this app

you may find it just as entertaining

now when you know where to look for the

decent text to 3D AI let me show you how

to clean the models a little bit and 3D

print them

I will be using blender to do this as

it's free and easy to follow

[Music]

open new blender file and delete

everything in your scene

now go to file

import glb

find the model you just downloaded and

import it

as you can see the mesh is a little torn

apart but we can fix it

first select the main body and click tab

now select any point on the surface and

click Ctrl l

this will highlight all the connected

vertices

Ctrl I to invert the selection and

delete vertices

that's how we get rid of all the

separated parts

tab again to exit edit mode and now

click on this little wrench

from the modifiers list select remesh

and play a little with the voxel size

the smaller the volume the more detailed

your object will be but show some mercy

to your computer don't push that slider

too hard

if your objects has some huge gaps like

my flower pot here try to fill the holes

with some basic shapes before you add

rimage modifier

a to add new object G to move it around

a bit

you don't have to be very precise with

it just cover the holes with a new shape

select the new object then your model

while holding shift and hit Ctrl J

now when you image it the new filling

will be nicely integrated into your

asset you can also smoothen the model a

little bit using subdivision surface

modifier

choose the amount that fits your asset

nicely

remember to apply all the modifiers and

you're good to go

select the model go to file export STL

and check the selection only box

save it wherever you want and open it in

your slicer and that's it this is how

you can 3D print any model designed for

you by an AI

the models generated by Mirage are quite

basic but they can be very helpful

addition to your 3D modeling routine

you can get a simple shape in a couple

of minutes and then adjust it yourself

the way you like

also if you are fine with their

minimalistic shapes and you don't need

complicated details you can print them

right away

[Music]

for me just the concept of those models

being made by artificial intelligence

makes them cool enough to have them

printed

compared with how easy and user-friendly

this whole process actually is I can

definitely consider Mirage as the best

text to 3D tool I've seen so far

like I said before this technology is

still being developed and we can all

expect better and better results as days

go by

but even though the generated assets are

quite basic this is actually a very

decent starting point

I'm super excited to watch the

artificial intelligence progress in 3D

modeling and I promise to keep you

updated on any relevant news

let me know in the comments about your

opinion on this topic and if you have

any other experience with the text to 3D

AI you're more than welcome to share it

I also invite you to check my other

videos about 3D modeling 3D printing

design and architecture if there are

some particular topics you find

interesting please mention them in the

comments I would be happy to know what

content would you like to see

thanks for watching and see you next

time

[Music]