Creating a 3D Character from Photos and Text
The first step of working with Animatic is getting a volumetric model of the character. You upload an image, and the program builds the three-dimensional geometry and textures. After that, the model can be checked, fitted onto a skeleton and animated.
In this article we look at how the process works, how to prepare the source image and what to pay attention to so that the result is clean.
What You Get as a Result
Generation creates a three-dimensional model with geometry — a surface you can rotate and view from any side. Together with the geometry, textures appear: colors, relief and surface characteristics that make the character resemble the source image.
The model is saved into the project folder and opens right away in the built-in viewer. It can then be passed on to the next stages: rigging, skinning and animation.
Which Image to Choose
The quality of the model depends directly on the source. Here is what helps achieve a good result:
- One character in the frame. Extra people and objects prevent the program from isolating the subject correctly.
- The whole character visible. If only the head and shoulders are visible, the legs and arms have to be inferred, and the result will be less accurate.
- Neutral background. A plain background makes it easier to separate the character from the surroundings.
- Even lighting. Harsh shadows and glare distort the texture colors: the program may take a shadow for part of the relief.
- Minimum occlusion. If an arm covers the torso or the legs are crossed, the hidden parts are hard to reconstruct.
- High resolution. Crisp details of the face, clothing and accessories transfer into the model better.
Text as a Starting Point
Besides an image, you can start from a text description: for example, "a warrior in heavy armor" or "a girl in a long cloak". The text defines the general look, and further details can be refined by preparing a matching image.
Choosing a Mode for Your Hardware
Generating 3D models requires video memory, and its size determines the available quality options:
- Basic mode suits GTX 1070-class cards with 8 GB of memory. A fast, economical option for your first experiments.
- Standard mode is for RTX 3060 cards with 12 GB of memory and above. The models come out more detailed, and the textures reproduce materials more accurately.
- Maximum mode targets the RTX 4090 and newer with 24 GB of memory and above. The highest quality, suitable for final models.
The program detects your computer's specs on its own and offers the available options. If there is not enough video memory, the mode switches to a lighter one and you still get a result without errors.
Step-by-Step Process
- Place the character image in the project folder inputs/[name]/imgs/.
- Open the program and select the character from the list.
- Start the 3D model creation and wait for it to finish. On weaker hardware this may take a few minutes.
- Open the result in the viewer and inspect the model from all sides.
- If you don't like the model, fix the source image and run the creation again.
How to Check the Result
Before further work, look at the model from several angles. Pay attention to:
- surface integrity: no holes, detached pieces or blobs;
- proportions: do the arms, legs and head match the source image;
- texture colors: no splotches inherited from shadows and glare.
If the model looks good, move on to the next step. Geometry errors are better fixed right away: later they are harder to notice, and the skeleton and animation would be built on a wrong basis.
What to Do Next
A finished model needs to be prepared for animation: build the skeleton and distribute the weights. This process is described in the article "Auto-Rigging a Character".