Text-to-3D: AI Image Converter — a New App by MagiScan
The Text-to-3D app is available on both iOS and Android. Today, we’ll review its free iOS version, which allows users to generate 3D models in three attempts and exporting in MP4, GLB, GLTF, STL, OBJ, USDZ or webURLing file formats.
The Philosophical Concept of the App
The core idea of the app is simple: you write a text, and it gets converted into a 3D model. The app first generates a 2D image (you can change the proposed images based on your text input) and then a 3D model.
It’s worth noting that the more detailed your description, the more refined and accurate the result will be. In the premium version, the process is faster (there’s a timer counting down), while in the free version, it’s slower.
The deeper philosophy behind Text-to-3D is that it’s not just a utility but a powerful tool for transforming reality into pure creativity. The ability to integrate 3D objects into the real world via AR opens up exciting prospects for photographers and artists. For example, street photographers can not only capture reality but also enrich their work with augmented objects, creating unique images. This blend of the physical and virtual worlds could become a new direction in art. I can already imagine many artists and photographers creating new movements — similar to how video game photography became a genre.
Occam’s Razor and UI/UX in Text to 3D
The first thing you notice is that the app’s interface is reminiscent of PolyCam — a flagship 3D scanning tool. While the idea of converting text into 3D models is not new, this app stands out for its simple and intuitive interface. We know many apps suffer from cliched designs, some even bordering on kitsch and the principle of “UI/UX done half-heartedly.” With Text-to-3D, it’s a different story.
It’s perfectly suited for children and teenagers due to its simplicity, but at the same time, it offers enough functionality to engage adult users. At the bottom, we see a panel with only three actions: generating a 3D model (either by writing text or attaching your own image), an archive where your models are stored, and a profile (where you can upgrade your subscription).
Nothing extra.
The app is so intuitive that the absence of a manual can be considered a sign of excellent design — if a child can use it, it’s designed right. This principle is comparable to explaining complex ideas: if you can’t explain them simply, you don’t fully understand them yourself. Therefore, the UI/UX of the application deserves a 10 out of 10.
Option 1: Cooking broccoli or converting a text in 3D
Testing the app on different objects produced mixed results. For example, generating a 3D model from the Russian word “broccoli” was successful — the model was clearly generated and could be easily integrated into augmented reality.
It’s worth noting that the app excellently detects surfaces, which is impressive since many AR apps struggle with this. This is due to both the app itself and the LiDAR on the iPhone. It will be interesting to test this on Android, where AR models work differently.
New Balance Sneakers: Detailing Issues
Testing the generation of a 3D model for a New Balance sneaker also revealed some amusing moments. The app generated only the right shoe, which was strange considering the request was for the whole pair. This may be due to the app’s interpretation of the text or image, not recognizing the pair as a single object. Another issue was with the back of the sneaker, which was poorly detailed. This could be related to the mesh reconstruction algorithms. During the texture conversion or segmentation process, the back part of the object may have been lost or smoothed, resulting in a blurry outcome. Neural networks responsible for processing 3D objects sometimes “smooth” or skip parts of the image if they are not clearly defined or lack contrast.
Option 2: Attach a 2D Image and Get a 3D Model
I also tested the image-adding feature by downloading three different images from three different sources: a goat (from Google Images), a toy Batman (from a toy store website), and an octopus. Batman and the octopus turned out well (as seen in the images), but the goat had issues. This could be due to the image quality and the difficulty in recognizing the figure and background.
A Genie in a Bottle, or the Business Model
For now, the business model is simple and user-friendly: no hard paywalls or “3 days for free”. Right from the start, you can generate whatever you want, but like a fairy tale, you only get three attempts. The business model is built around four types of subscriptions: weekly, monthly, yearly (allowing you to save), and a one-time lifetime subscription. Prices can be seen in the image.
Some recommendations
I also liked that you can report an issue in the app and like in a movie, I have found a bug: when choosing Polish złoty as the currency, the subscription price was displayed incorrectly. This needs fixing. Overall, everything works great.
Conclusion
Despite some flaws, Text to 3D has significant potential. Issues with object segmentation, like with New Balance sneakers, can be resolved with algorithm improvements. The app’s rating is 8+ out of 10, and we can expect the developers to address these current issues and release an improved version in the future!
Try it out: https://apps.apple.com/us/app/textto3d-ai-image-converter/id6572296572
