Digital characters and avatars appeared even before the computer and console graphics. I remember the text-based version of Pac-Man available at the ancient DVK-4 PC. Later, some plausible characters appeared in the movies. Today, the time has come for the photorealistic characters operating in (almost) real-time. The tech, running in the cloud on high-performance systems, is nearly ready to step up our game.

If you need a customized and augmented digital human for the app, game, or simulation, there are various platforms and tools that can help you. Let’s explore them, highlighting some of their features.

7 best digital human providers

1. MetaHuman Creator is the technology launched by Epic Games. As of January 2022, it is in early access and will probably be released with Unreal Engine 5. You can use the online tool that streams results to your local computer, making models from the presets and importing them into Unreal Engine. Such a pipeline is excellent for creating apps and games.
2. Soul Machines is the studio for creating an avatar or chatbot on demand. The preview results are impressive, but we haven't had a chance to get our hands on technology under the hood.
3. Hologress. This digital human platform is positioned as the first interactive virtual closet is also in early access. You can dress up your characters using various presets.
4. Human Generator V3. Nearly everything for Blender is free, and surprisingly, this piece of technology is not. From the visual standpoint, the digital humans generated with this plugin are less detailed than the platforms on the list.
5. Character Creator 3. We've come across this tech at the GTC sessions on Nvidia's “metaverse for engineers” named Omniverse. The app comes with presets for almost everything, including makeup. The suite includes an animation package and the export works for Omniverse and Unreal Engine. The app limits the models' usage with basic license-generating images, and videos are royalty-free. The solution is perfect for data generation.
6. Datagen is a company that generates datasets on demand. The results can be considered hyper-realistic, and therefore a bit spooky.
7. Ziva Real-Time Face Trainer – the tech is aimed mainly at creating facial animations. The Unforgettable Characters do look unforgettable and come with an ML-trained automated facial rig.

To summarize the information about various digital human providers, we assigned scores to them, according to the different criteria, and created a comparison table.

Notes to the comparison table:

Price (0-3): 0 – the price is not disclosed; 1 – the price is >$1000; 2 – the price is <$1000; 3 - the tool is free.

License (0-3): 0 – data about the licenses is not available, 1 – the license is paid with some limitations, 2 – the license is free with some limitations (digital human models, AI training) 3 – the license is free.

Photoreal visuals (0-1): 0 – the visuals are not photo-real, 1 – the visuals are photo-real.

RAW Data (0-1): 0 – doesn’t have raw data access; 1 - has raw data access via fbx or USD

USD support (0-1): 0 – doesn’t support USD; 1 – supports USD for Omniverse.

How SoftServe generates custom digital human datasets on-demand?

Whatever digital human provider you choose, you might discover that collecting quality training data takes significant time.

At SoftServe, we sought an effective solution to this problem. In the last several years, we’ve developed a pipeline to simplify the generation of custom 2D datasets and 3D digital human models.

Initially, we experimented with creating Digital Data Generation Pipeline for much less complicated models than humans. Ironically, it turned out that license limitations restrict to use MetaHuman Creator for these purposes. Then we've looked for the other options and found out that Character Creator has the opposite limitations: you can use images and can't use models with the default license. MetaHuman Creator and Character Creator 3 seem to cover the most common cases for Digital Humans data generation and app development. A testament to our advancements in this domain, we have created immersive experience with a digital human named Alnair.

Today, we create datasets for hundreds of thousands of randomized samples, all labeled with accurate data for ML training workflows. The process requires much less generation and delivery time than collecting image datasets manually.

We deliver a tailored digital human dataset in several weeks. As ML becomes more sophisticated in algorithms, we take special care of the data accurateness to accelerate and improve the quality of final AI or computer vision models.

LET’S TALK about how SoftServe can generate and provide you with nearly perfect metadata to elevate your digital human experience.