Experience Imagine 3D firsthand.

Share this story

Imagine 3D by Luma AI: A Unique Approach to Text-to-3D Generation

Today, we are exploring Luma AI’s Imagine 3D, a groundbreaking text-to-3D service. The concept is simple – you input a text prompt, and the AI generates a 3D model based on that prompt. While text to image AI generators are quite popular, Imagine 3D takes it a step further by creating fully realized 3D models instead of just images. However, it does face some similar challenges, as we will discover.

Luma AI, previously known as Luma Labs, recently caught attention with its innovative smartphone 3D scanning technology called NeRF (Neural Radiance Fields). NeRF is an AI-powered system that takes multiple images of an object and generates new images from alternate angles. Building upon this foundation, Luma AI developed an effective 3D scanning system that has the potential to surpass traditional photogrammetry as the leading non-hardware 3D scanning approach. By combining NeRF technology with an advanced modeling system, Luma AI aimed to revolutionize the way 3D models are generated.

Curious about its capabilities, I decided to give Imagine 3D a try. However, it is important to note that Imagine 3D is currently in alpha testing, meaning it is still in the early stages of development. Knowing this, I was prepared for a less-than-perfect experience. To access the system, you need an ID on Luma AI’s platform, which can be obtained easily by signing up with Apple or Google authentication. Surprisingly, there’s no option to sign up using just an email address. After gaining access to the scanning app, you must apply for the Imagine 3D waitlist. I applied several months ago but never heard back. Eventually, I discovered that joining the company’s Discord server, navigating to the Imagine 3D channel, and requesting activation was necessary. To my surprise, my activation request was granted almost immediately, leaving me wondering why I hadn’t tried this earlier.

Using Imagine 3D is incredibly straightforward, and that’s precisely the intention: simplifying the process of building a 3D model without requiring extensive CAD work. As displayed above, there’s a simple text input box where you can type in your prompt. It seems that the AI model has its limitations, as the range of possibilities appears to be somewhat narrow. Once you hit the generate button, you patiently wait for the results to appear. Typically, the generation process takes less than an hour, and it’s even possible to have multiple generations in progress simultaneously. Once a generation is complete, you are presented with a fully rotatable 3D view of the model. In terms of 3D printing, you have the option to download the generated model in .GLTF, .USDZ, or .OBJ formats. For my testing purposes, I used the .OBJ format as it was the most compatible format with my 3D tools. I was pleasantly surprised to find that the model came with textures. Furthermore, it’s worth noting that the generated models are in full color. To access the downloaded model, you must unzip the file and use a 3D tool to open the relevant file format. In my case, I utilized MeshLab to open the .OBJ file.

Now that we’ve covered the basics of Imagine 3D, let’s dive into the results. I tested the system with various prompts, ranging from simple to complex. The results were undoubtedly intriguing, albeit not flawless. Let’s examine a few examples. When prompted with “marble bust of Winston Churchill with a cigar,” the generated model looked quite accurate, except for the missing cigar. This led me to speculate that Imagine 3D encounters challenges when it comes to combining multiple elements. To further explore this, I tried the prompt “a happy elephant with angel wings.” At first glance, the model seemed promising, but upon closer inspection, I discovered a nightmarish surprise – instead of angel wings, the elephant had two trunks and three eyes, including a freakish cyclops eye between the trunks. Quite an unexpected result! Lastly, I ventured to create “a cowboy riding a giraffe,” resulting in the following model: [model not shown].

In conclusion, Imagine 3D is an exciting concept with considerable potential. The ease of use, coupled with the ability to generate 3D models from text prompts, makes it an intriguing tool for various applications. However, it’s important to note that the system is still in its early stages, and the generated models may not always align with the intended prompts. Nevertheless, with further development and fine-tuning, Imagine 3D could revolutionize the way 3D models are created, opening up new possibilities for creativity and design.

Title: The Transformative Power of Imagine 3D: A New Frontier in 3D Modeling

We have all heard the saying, “Seeing is believing,” and I had a chance to experience this firsthand with Imagine 3D. At first glance, it seemed like an ordinary 3D modeling tool, until I stumbled upon something extraordinary – our fellow participant had three arms! Well, not exactly, but let me explain.

Imagine 3D offers users the ability to create 3D models by inputting specific requests. Intrigued by its capabilities, I decided to put it to the test. I requested the creation of two separate objects – a polar bear and a pineapple. Initially, it seemed like Imagine 3D could indeed meet my expectations as the two objects materialized. However, my excitement was short-lived because when I rotated the 3D model, I discovered that the system had merged the two concepts into one mishmash of a hybrid. Instead of two distinct objects, I was left with an unusual combination.

This incident got my gears turning. I wondered if Imagine 3D would fare better with abstract concepts. To find out, I requested the creation of “The essence of life.” The result was intriguing, but puzzling. The 3D model seemed to be an extruded image of a flower, leaving me wondering how it related to the essence of life. Was the flower a symbolic representation? Perhaps, but it did not align with my original intention.

As evident from these examples, it became clear that Imagine 3D possesses limitations when it comes to generating specific 3D models, especially with precise measurements or intricate details. It appears to be more inclined towards a 2D style approach, where the created model matches the prompt only from specific angles. Nonetheless, this experience also revealed the system’s capacity to produce unexpected 3D models that are actually printable. Nevertheless, brace yourself for a reliance on support materials, as flat surfaces are virtually non-existent.

What truly fascinated me about Imagine 3D is the glimpse it provides into the future of 3D modeling. Instead of spending hours immersed in a complex 3D CAD system to create simple objects, Imagine 3D offers the tantalizing prospect of merely requesting them. While the current version of Imagine 3D may still be rough around the edges, its rapid emergence in just a matter of months is a testament to its potential. It represents the beginning of a new era in 3D modeling.

Ultimately, Imagine 3D serves as a living example of what lies ahead for the 3D industry. It may not yet be fully refined, but it offers a tantalizing vision of what the future holds. Let us embrace the unknown and eagerly anticipate where this technology will take us in the coming years. After all, the journey to a fully-realized 3D future has only just begun.

Via Imagine 3D.

Share this post:

Original source


Share this story

Leave a Reply

Your email address will not be published. Required fields are marked *