From 2D Sketches to 3D Figures|watabin|愛(AI)と共に

From 2D Sketches to 3D Figures|watabin|愛(AI)と共に

It seems we have entered an era where you can create 3D figures from just a single photo or sketch.

3D conversion tools and generative AI for 3D have also been provided as production tools.

If you use generative AI, you can create data by giving it a prompt.
It felt like you could create high-precision works if you had multiple target images.

By the way, to create good 3D data, you must first remove unnecessary information like the background from the 2D data!
On an iPhone, you can easily remove photo backgrounds using only standard features.

The example is a sketch, but I might want to turn a generative AI image or a photo into a figure😅

You can create a figure by outputting STL data, but STL data can also be used for things like 3D game characters.

Also, if you put photos or rough sketches into generative AI and ask it to arrange them in a “Ghibli style” or “Disney style,” it will convert the image to look like that.

Additionally, if you put multiple images into an AI and ask it to “make the features of this image into a prompt,” it seems that if there is a function to create 3D data from text, it will create data that is quite close to the image you had in mind.

Technology that generates 3D models or 360-degree viewable solid data from 2D data (a single photo, illustration, hand-drawn rough sketch, etc.) has become widely available to the public in recent years due to the rapid evolution of AI.

The phenomenon often seen on social media where “a 3D object is created and moves from a single picture” is mainly done using tools and technologies called “Image-to-3D AI.”

Below is an overview of how it works and some representative tools that are currently commonly used.

1. How 2D to 3D/360-degree images are created

Hi3D

AI uses the outlines, shading, and perspective of the input 2D image (photo or sketch) as clues to highly predict and fill in “what the back side looks like (the shape of the hidden parts)”.

  • Shape (geometry) generation: Automatically builds a 3D mesh (framework) of the object. Meshy AI

  • Texture (texture/color) generation: Applies surface colors and textures, finishing it so that it does not break down when viewed from any angle (360 degrees). Sloyd

  • Rigging/Animation: For characters, some tools have the function to automatically insert a skeleton (rig) to make them movable. Kimi.ai

2. Recommended representative tools

Depending on the purpose and what you want to create, the following browser-based and cloud-based AI tools are commonly used.

① Meshy AI

  • Features: A versatile tool that can generate high-quality, textured 3D models in about one minute from photos, illustrations, rough sketches, and even text.

  • Strengths: It has extremely high shape reproducibility and is equipped with “Smart Topology” to optimize polygon counts for games, automatic part segmentation, and AI animation features. It supports a wide range of outputs, including FBX, OBJ, GLB, and STL formats for 3D printing.

Meshy AI

② Tripo AI

  • Features: A 3D generation platform that is easy for beginners to use, with overwhelming generation speed as its selling point.

  • Strengths: You can quickly create drafts from a single image or sketch, and easily turn everything from anime-style characters to small items into 3D objects.

  • 3D Printer and 1 other item

Tripo

③ Luma AI / Genie

  • Features: A tool that excels at creating realistic 3D assets from photos taken by circling an object with a smartphone or from a single image.

  • Strengths: It has a reputation for high-quality representation of photo-based objects, product 3D modeling, and AR/Web applications.

Luma AI (a convenient service that can create videos from text or images)
Genie 4: The latest world model announced by Google.

④ Rodin / Hunyuan3D (Tencent-affiliated, etc.)

  • Features: A high-precision AI model specialized in beautifully reproducing figures, characters, and anime-style designs.

  • Strengths: It is chosen when you want to focus on the quality of details (such as the wrinkles in clothing or the three-dimensionality of hair).

  • 3D Printer

Rodin
Hunyuan 3D

3. Tips for creating successfully from sketches or photos

To produce high-quality models like those you see on social media, there are a few tricks to the 2D data you input.

  • Keep the background simple: Using images with clear boundaries between the subject and the background, such as white backgrounds or transparent PNGs, makes it harder for the AI to misinterpret the shape.

  • Make lines clear in rough sketches: If lines overlap too much, the AI will get confused, so using clean line art or adding light shading (shadow colors) will improve depth accuracy.

  • Utilize multiple angles (multi-view): If possible, uploading several images drawn (or photographed) not just from the front but also from the side or at an angle will make the shape of the back more accurate (many tools support uploading multiple images at once).

First, try dropping a photo or illustration you have into an “Image-to-3D” feature like Meshy AI or Tripo AI, which you can try for free in your browser, and you will be surprised at how easy it is.

Comparison of recommended 3D model generation AI by use case

There are various types of 3D model generation AI, each with different strengths and features.
First, let’s look at the major services in a comparison table to find the tool that suits you.

Comparison of recommended 3D model generation AI by use case

When choosing a 3D model generation AI, it is easier to choose if you decide on your purpose: what kind of 3D model you want to make and what you want to use it for after it is created.

3D data creation

There are convenient services and workflows that allow you to create 3D models from 2D data (photos or illustrations) using AI and then directly connect them to 3D printer output or figure creation.

Broadly speaking, there are two approaches: “(1) Tools that have 3D printing features or slicer integration on the AI tool side”, and “(2) Tools where you create the data and leave it to a print-on-demand service.”

AI tools with 3D printing integration and optimization features

The representative AI tools mentioned above are equipped with features to convert files into formats for 3D printing (STL or 3MF formats) and optimization features to prevent printing errors (such as hole filling and thickness adjustment).

(1) Meshy AI + Slicer integration

  • Features: You can download 3D models generated from images or text directly in output formats for 3D printers (STL / 3MF).

  • Strengths (Direct Slicer Integration): It features direct sending (one-click integration) to popular 3D printing software (slicers) such as Bambu Studio, OrcaSlicer, and Creality Print. This significantly reduces the effort from AI generation to print preparation (slicing).

② Tripo AI

  • Features: You can quickly generate 3D models from photos or illustrations and export them as data for 3D printers (STL, etc.).

  • Strengths: It includes a feature that automatically cleans up “holes” and “non-manifold edges” (broken surfaces) that often cause errors when 3D printing, making it relatively safe even for beginners to create models for FDM (Fused Deposition Modeling) or SLA (resin) printers.

2. General workflow for connecting “data creation to figure formation”

If you do not have a 3D printer at home, or if you want to create a high-quality professional figure (especially one with the smooth, beautiful texture unique to resin), you can combine services using the following workflow.

  1. Convert to 3D data (STL format) with AI

  2. Fine-tuning shape and size (as needed)

  3. Printing (Self-made or outsourced)

    • If printing at home: Using a resin 3D printer (a type that uses light-curing resin to create detailed figures beautifully) makes it easier to reproduce precise shapes like those of anime figures.

    • If using an outsourced printing service: By uploading STL data to an online modeling service that outputs 3D data for individuals (such as “DMM.make 3D Print”), you can choose the material (acrylic resin, full-color gypsum, etc.) and have the physical figure mailed to your home.

Points and tips for creating figures

  • The “thickness” of hair and clothing: Depending on the photo or illustration, the tips of hair may become thin, flat plates that cannot be output by a 3D printer. A tip for clean output is to “give it some thickness/solidity” on the AI tool side or in the slicer.

  • How to make it stand on its own: If a character has unstable feet, the data generated by AI as-is will not stand on its own. Attaching a disc-shaped “base” to the bottom in slicer software before outputting will allow it to be displayed stably as a figure.

The best shortcut is to first use the free tiers of Meshy AI or Tripo AI to turn illustrations or photos into 3D objects, download them in STL format, and take a look at “what kind of data it becomes.”

💰Production costs when using Meshy AI

The costs involved in creating a figure by combining Meshy AI and slicer software (such as Bambu Studio or OrcaSlicer) are divided into two parts: “(1) AI tool usage fees”and“(2) Actual printing costs on the 3D printer side.”

1. Meshy AI usage costs

You can try Meshy AI starting with the free plan, but if you are creating data for figures in earnest (for commercial use or high-resolution output), you should consider a paid plan.

2. Costs for slicer software and printers

Slicer software (such as Bambu Studio or OrcaSlicer) can be downloaded and used completely for free. The actual cost of printing varies significantly depending on the output method.

Pattern A: Printing with a home 3D printer (cost-effective in the medium to long term)

For figure production, a resin (SLA/DLP) 3D printer is recommended as it produces fine details beautifully.

  • Initial costs (hardware): The printer itself, specialized resin, and cleaning/curing UV equipment will cost approximately 30,000 to 60,000 yen upfront.

  • Running costs (per figure):

    • Resin costs about 3,000 to 5,000 yen per bottle (1kg).

    • For a standard palm-sized figure about 10cm tall, the resin cost is only about 100 to 300 yen, making mass production very affordable.

Pattern B: Using a print-on-demand service (zero initial cost)

If you cannot keep a printer at home, you can upload the STL data generated by Meshy AI to a service like “DMM.make 3D Print” to have it printed for you.

Estimated total budget

  • For a trial run (free AI + free slicer):

  • For making figure creation a serious hobby at home:

    • Initial investment (printer, etc.) around 40,000 yen + Meshy Pro fee (approx. 3,000 yen/month) + resin cost (several hundred yen per figure)

  • For creating high-quality figures on a spot basis (1-2 items):

    • Meshy Pro fee (approx. 3,000 yen/month, or free tier) + printing service fees (several thousand yen per figure)

Similar Posts

Leave a Reply

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