2024 AI in Gaming Industry

Industrial Trend

According to Global Gaming Industry Trend, KOCCA selected Gen AI as the main industrial trends. Gaming has been shown a high aboptability on new technologies in order to achieve developments and innovations so far. Gen AI will be the an major tech that increate the productivity that helps creating assets such as image, video, game, and storyline.

More than half of the video game development process will be supported by generative AI within the next five to 10 years, according to new research by Bain & Company.

Company Research

Krafton

Krafton is a South Korean video game publisher based in Seoul. Krafton is known as a publisher of Pubg(PlayerUnknown’s Battlegrounds)

Research Area

Vision & Animation

Vision & animation technology generates visual assets or objects in 2D or 3D form. Krafton’s focus in this field is to interpret and understand images or videso, then convert them into various formats for application.

For example, traftons study domains such as video-to-motion - where they build 3D models from 2D images or transfer human movement in a video to a target object - or lip syncing, where they generates lip animations that correspond to the input text.

KRAFTON AI’s goal is to use these studies to streamline processes that require designers or animators to execute them manually, while at the same time maintaining realistic quality.

Voice Synthesis

Voice synthesis technology allows computers to imitate human speech and speak like humans. Our focus in this field is to mix emotion into natural, human-like voices.

For example, we study how to add emotions and tone to text-to-speech (TTS) output, where TTS is a technology that converts input sentences to human voices. We also study voice conversion, which transforms an input voice into a different type of voice.

KRAFTON AI uses voice technology to develop innovative solutions that can express rich emotion, tone, and accents. Our goal is to transcend physical limitations that might be inherent in speech to expand the potential of communication.

Multi-modal Learning

Multi-modal learning is the ability of a model to understand and process different types of data at once. Our focus in this field is to use a mix of data containing various formats such as images, text, audio, and videos to its fullest capacity by leveraging the unique information that each data format contains.

For example, we could build large-scale databases that contain both images and text for specific topics, or link language models to TTS technology so we can generate audio that fits the context and meaning of what is to be said.

KRAFTON AI no longer classifies deep learning applications into individual fields, but rather tries to take a more general approach in solving more complicated and difficult problems. We must fuse together technologies spanning all fields to make possible the interaction between humans and computers that we envision.

Language Model

Language models help computers understand and generate text in human languages. Our focus in this field is to solve specific problems using the many applications that emerged with the introduction of foundational models. For example, we could build a casual, conversational system using GPT-based models or build a Q&A system specialized in a specific field. KRAFTON AI tries to improve the capacity of computers to understand human language, a crucial part of computer-human interaction, and build an experience that allows the user to forge deeper relationships with the system.

Reinforcement learning

Reinforcement learning is a technique where an agent is developed to interact with its environment so it can learn how to find the optimal action strategy (policy).

This technology is closest to KRAFTON’s core drive as a game developer and maker of masterpieces. For example, we could train the agent on important rules and strategies from a game and have it compete with other users or help developers test the game as it is being designed.

Reinforcement learning will play a crucial role in interactions between humans and computers as it helps AI develop the human-like capacity to learn and make decisions about complicated problems.

Data-centric AI

Data-centric AI focus on the quality of data used to enhance model performance. We study data management strategies surrounding data collection, processing, labeling, augmentation, and other processes that play an important role in building high-quality datasets.

For example, we study whether it is possible to achieve the same level of performance with less data or what type of data we would need to supplement for improved model performance.

Data-centric research based on deep learning enables AI models to make more accurate and powerful predictions. This field contributes to improving the reliability and fairness of AIs and is crucial in establishing a long-term direction and foundation for research organizations.

Application

Virtual Friend

Creating agents in various forms of friends who can play games with users like you. Those agents can understand the current situation in the game, chat, and play with you.

Singing Voice Synthesis

Creating songs by applying the unique voice and pitch of the character to the melody and lyrics. We have produced a song for ANA, one of KRAFTON’s Virtual Humans.

3D Avatar Generation

Generating 3D avatars trained in specific styles using photos of actual people as input. We can also create new styles by adding data.

In-house Application

Creating various tools that can help with game development and distribute them throughout KRAFTON. In this process, we aim to discover the potential to bring innovation to the gaming industry as a whole.

Application

NVIDIA ACE

NVIDIA ACE is a suite of technologies for bringing digital humans, AI non-player characters (NPCs), and interactive avatars to life with generative AI.

AI technology

RAG

Retrieval-augmented generation (RAG) is a technique for enhancing the accuracy and reliability of generative AI models with facts fetched from external sources.

In other words, it fills a gap in how LLMs work. Under the hood, LLMs are neural networks, typically measured by how many parameters they contain. An LLM’s parameters essentially represent the general patterns of how humans use words to form sentences.

That deep understanding, sometimes called parameterized knowledge, makes LLMs useful in responding to general prompts at light speed. However, it does not serve users who want a deeper dive into a current or more specific topic.