Make-A-Character: Text-to-3D Human Generation 129

Make-A-Character allows you to efficiently and at scale generate hyper-realistic 3Ddigital humans with full-stack artistic capabilities by describing a person's face shape, facial features, and hairstyle in a single sentence. The generated results include a head and body, accessories and clothing, and support being driven by facial animation assets, bringing the 3D digital human to life with vivid dynamic display.You can also customize facial features such as face shape, eye color, hairstyle, eyebrow type, mouth and nose, as well as adding wrinkles and freckles.Make-A-Charactersupports both English and Chinese prompts.

Make-A-Character Project URL:https://human3daigc.github.io/MACH/

Make-A-Character Demo URL:https://www.modelscope.cn/studios/XR-3D/InstructDynamicAvatar/summary

Main Features of Make-A-Character:

1. Highly Realistic: Characters are generated based on a dataset of real human scans, with hairstyles using actual hair strands rather than meshes, and rendered using Unreal Engine's Physically Based Rendering (PBR) technology for high-quality real-time rendering.

2. Complete Models: Every character is a full model, including eyes, tongue, teeth, full body, and clothing, ready for immediate use without additional modeling.

3. Animatable: Characters are equipped with a complex skeletal rig that supports standard animation, enhancing their realistic appearance and enabling diverse applications.

4. Controllable: Users can customize facial features in detail, such as face shape, eye color, hairstyle, eyebrow type, mouth and nose, as well as adding wrinkles and freckles.

5. Industry Compatible: The generated 3D characters can be seamlessly integrated into existing computer graphics (CG) workflows, particularly those commonly used in the gaming and film industries.

How Make-A-Character Works:

Make-A-Character (MACH) combines large language models, vision models, and 3D generation techniques.

The framework leverages a large language model (LLM) to extract various facial attributes (e.g., face shape, eye shape, mouth shape, hairstyle and color, glasses type). These semantic attributes are then mapped to corresponding visual cues, which in turn guide the generation of a reference portrait image using Stable Diffusion together with ControlNet. Through a series of 2D face parsing and 3D generation modules, the mesh and texture of the target face are produced and assembled along with other matching accessories. The parametric representation enables easy animation of the generated 3D avatar.

How to Use Make-A-Character:

1. Enter a face description

2. Click Generate and wait for 2 minutes

3. View the generated result

Through these steps, Make-A-Character can quickly generate realistic, complete, and animatable 3D characters from simple text descriptions, suitable for a variety of entertainment and professional scenarios.