1. Introduction
Although it is gradually emerging from the sluggish atmosphere caused by the recent COVID-19 pandemic, the economy is still experiencing a slowdown in vitality in 2023 due to global low growth and low youth employment rates. In this situation, new drivers of leap forward and regeneration are needed, and new industries and creative and innovative companies are required as well as maintaining the competitiveness of existing major industries and leading companies. In this context, we should pay attention to the content industry. Recently, in modern society, the form of media consumption has changed very rapidly, and after the COVID-19 pandemic, SNS media usage has soared and information storms have accelerated, and videos that must be interested in a short period of time are preferred. Producers should be able to supply various media contents such as videos, webtoons, and graphics more quickly in line with this trend. However, content production is relatively time-consuming and expensive because it goes through the process of drawing and assembling images on a frame-by-frame basis, making it difficult for content consumers to quickly supply content. Due to these times and realistic constraints, content creators need to consider new strategies. With the recent rapid evolution of image generation AI technology, the content industry has announced R&D results and commercial content production processes, showing that it can be fully utilized in the production pipeline and that the capabilities required by existing personnel are changing. Experts say it won’t be long before the era of full-fledged changes in content production pipelines using AI technology comes1). Through the content industry, we can bring about a new leap forward in the economy from a national crisis. In particular, the spread of global OTT services and the enormousity of the YouTube content market have brought an important turning point in the content industry, and it is becoming a major industry in Korea. The content industry is being reborn as a content industry through a huge platform by converging with domestic and foreign innovative technologies such as AI, big data, 5G/6G communication, and metaverse, which converts our lives to digital and accounts for a large portion of industrial demand.
Accordingly, university education needs to play a role in providing a supply for this demand. AI convergence content education is already seen as not a technology for the future, but for the present. Therefore, in order to respond appropriately to the future, it is necessary to establish not only content creation and production capabilities, but also core capabilities to play technical and management roles and industry-academic curriculum to respond to efficient content development in the new technology industry. To this end, it is necessary to introduce a Department of Contents, University Curriculum that can lead the direction and speed of change in the AI content industry, and it is important to cultivate the ability to take the lead in responding to industrial changes and acting as a supplier to demand. Through this study, it is expected that it will lead to changes in the industry and lead to the growth of the innovative content industry.
This study focuses on the popularization of image-generating AI such as stable diffusion, Midjourney, and Fire Fly, the need to apply AI future and curriculum, establishing content-related departments at universities that respond to the AI transformation era, and spreading AI content production processes. Until now, most AI-related departments and curriculums have been opened in engineering, but the AI revolution, called by image-generating AI such as Stable Diffusion and Midjourney, has expanded to humanities and arts, and most AI-related departments are selecting new students regardless of the field. The case of applying this in a content-related department and establishing it as a curriculum for production and planning is the AI design department of Kookmin University2). However, considering the current social utilization and speed of development of AI image-related tools, even if AI is not reflected in the name of the department, it is urgent to introduce it into the curriculum through re-establishment of talent. Therefore, this study proposes the direction of the application of the curriculum to create convergence capabilities, experience-oriented, and value with the aim of producing AI-adaptive content talent that can be quickly put into the AI content production process applied in industrial sites. In fact, various content fields have recently used AI to produce various animations, movies, and advertisements, which are helping companies improve their imagination and save time and money.3)
2. Contents and Methods of Research
2.1 The Contents of a Study
The current status and necessity of the image generation AI curriculum of domestic university content-related departments are as follows.
[research question 1]
What are the commercialized image generation AI platforms and principles?
[research question 2]
What are the image and video works produced with AI technology, and at what stage can they be applied to content production?
[research question 3]
What is the current status of AI content-related department openings and educational applications in Korea, and what is the application direction for each educational course according to the current level of development?
[research question 4]
What are the necessary competencies for training AI adaptive content production personnel?
2.2 Method of research
As a method of research, first, we summarize the current status and utilization of systems that can generate image and video content with current AI technology.
Second, we will examine the applicability of university education by analyzing cases of video content and animation works produced with AI technology and examining the extent to which the technology can be implemented at the current level of development.
Third, we look into the current status of AI convergence education in content-related departments at domestic university institutions.
For the current status of university institutions, we investigate the operation status of AI content-related departments / majors, and then design the ideal talent and curriculum for nurturing AI adaptive content talent. Through this, the direction of application of the relevant curriculum is derived.
3. Main Subject
3.1 Image Generation AI Platform and Principles
Image creation through AI is one of the fields that has made great progress in recent years. Deep learning technology is used to train artificial intelligence models to create new images. To achieve this, a model must be trained using a large amount of image data, and data collection is very important.
3.1.1 Creation principles and characteristics of AI image creation platform
Image creation platforms through AI are being developed and evolved in various ways, and representative tools include DALL-E and DALL-E mini made by Open AI, imagen and parti made by Google, GauGAN2 made by NVIDIA, and Midjourney made by Midjourney. There is. Above all, as Stability AI’s Stable Diffusion, which is suitable for creating cartoon or animation characters, was distributed under an open source license on August 22, 2022, the number of Image Generation AI users has increased exponentially, and various experiments and Contents began to boom. First, you can understand the creation principle by looking at the characteristics of NVIDIA’s StyleGAN, which was launched in the early days (2018) and continues to experiment and evolve as an image generation AI tool. StyleGAN is a deep learning model that generates high-quality images using generative adversarial network (GAN), an artificial intelligence technology.
StyleGAN, unlike existing GAN models, has the characteristic that the generated images are high resolution and high quality. Additionally, the generated images have a variety of styles and characteristics, and also show high creativity.
<Figure 1> Tero Karras et al. [2018: 2], Naver Wikipedia
There are two major versions of StyleGAN. The first version is StyleGAN, which generates the entire image, and the second version is StyleGAN2-ADA, which manipulates part of the image. Additionally, instead of using random noise during the image generation process, StyleGAN generates images using random style vectors, allowing various features of the image to be manipulated, and plays a major role in making the generated images more diverse and creative. Do it. By style mixing (by giving different style vectors for each resolution) the different w vectors used to create the three photos on the left, the photo on the right containing the styles of all three images is created [Karras and Aila, 2019]. Below is an example photo using StyleGAN.
<Figure 2> Example Photo Created Through StyleGAN
The second representative AI image creation platform is DALL-E. It is an AI image creation platform developed by OpenAI, and creates new images based on sentences entered in natural language. DALL-E has the feature of being able to generate high-resolution images from 256*256 pixels to 1024*1024 pixels, and has the ability to create better images through continuous learning. The model is continuously updated and learning is in progress using more data. Additionally, it can create various images depending on the sentence you enter. Below is an image created by entering the sentence “squirrel holding a cell phone” using DALL-E.
First, it is a deep learning algorithm. Most AI image creation platforms use deep learning algorithms to create images, which allows for the creation of more sophisticated and diverse images.
<Figure 3> Example photo created through DALL-E
These AI image creation platforms have something in common.
Second, high-resolution image creation. Many AI image creation platforms generate high-resolution images, allowing users to create more detailed and sophisticated images.
Third, scalability. The AI image creation platform is scalable so that it can be used in a variety of fields, allowing users to create and utilize images for various purposes.
Lastly, improvements and updates. Most AI image creation platforms continuously improve and update models, helping users achieve higher levels of image creation capabilities. Below is a table summarizing the commonalities of AI image creation platforms.
We will analyze the principles and characteristics of Stable Diffusion and Midjourney, which are rapidly rising after 2022 and hastening the possibility of AI use in the content industry.
First, Stable Diffusion is a text-to-image artificial intelligence model distributed by Stability AI under an open source license, https://namu.wiki/w/Stable%20Diffusion released on August 22, 2022.
<Table 1> Common Principles of AI Image Creation Platforms
It is a deep learning artificial intelligence model.
Unlike the existing text-to-image models analyzed previously, such as NVIDIA’s StyleGAN or OpenAI’s Dall-e 2, it was released with significantly reduced computer resources so that it can be run with less than 4GB of VRAM, and was recently released as a smartphone app that can be run on the iPhone. It was released. Unlike other AI platforms, stable diffusion can relatively maintain the initially designed form when the character moves and moves, and allows the user to implement images by adjusting detailed items numerically, making it more suitable for use in animation.
Next, Midjourney is artificial intelligence software developed by Midjourney. If you enter text in English in the prompt application window or create and insert an image as a link, the desired image will be created and is similar to DALL-E analyzed previously.
<Figure 4> Example Picture Created Through DALL-E4)
Midjourney takes place on a Discord server, and if you use the free plan, anyone can view your creations. If you don’t want this, you need a paid subscription.
Many created works are being produced and presented to the public in conjunction with ChatGPT, and the painting ‘Space Opera Theater’ produced with Midjourney in 2022 became famous by winning first place in the digital art category at the Colorado State Fair Art Competition.
<Figure 5> Midjourney Creation Work ‘Space Opera Theater’
The number of users is rapidly increasing to the extent that various video works and memes are being produced and released by creating desired images through input prompts in English and linking them with other AI programs such as D-ID. It is also widely used in commercial content.
<Figure 6> Hoseo University Animation Department AI Creation Lab AI Avatar Meme’ video produced with Midjourney [Park and Lee, 2023]
The characteristics of the two programs above are summarized in a table as follows.
<Table 2> Characteristics of Midjourney and Stable Diffusion
3.1.2 Utilization of the Image Generation AI Platform
Image creation technology through AI is used in various fields.
Firstly, it is the field of works of art. AI programs such as StyleGAN analyze artists’ works and create new works based on this. Additionally, in the design field, it can be used to create new design ideas. In fact, recently, AI-generated works have been exhibited at exhibitions, and some works have even been sold at auction at very high prices.5)
Second is the gaming and virtual field. In the field of using AI to generate graphics for games and virtual worlds, AI is used to create more realistic graphics.6)
Thirdly, it is the film and advertising industry. In the film and TV industry, AI is used to create new environments and special effects. For example, Weta Digital uses AI to generate various creatures and creatures in its movies7). Until now, most individuals have used AI or startups have created it to promote their own AI solutions, but it is true that commercial forms have been on the rise recently. Recently, advertisement videos for Samsung Life Insurance and McDonald’s were also produced, and the background music inserted in the advertisement was also created using AI, with subtitles explaining that it was produced using AI.
Fourth, the medical field. In the medical field, AI is used to generate new images or analyze images to diagnose diseases.8)
Below is a table summarizing the use of the AI image creation platform.
<Table 3> Utilization of AI Image Creation Platform
3.2 Production and Application of AI Content at this Stage
AI image generation technology provides various ways to assist creators and increase efficiency in the animation production process. In fact, the number of cases of animation production using AI image generation technology is increasing, presenting new possibilities in line with changes in the industrial environment. Image generation AI can create a variety of images depending on the diversity and quality of data, obtain unexpected results based on learned data, and provide new ideas to creators. In addition, recent image generation AI has improved its ability to generate images with high resolution and fine details due to the development of deep learning models and increased computing power, enabling the creation of high-quality images and personalized images according to user needs. Because it can be created, it is possible to provide targeted content in the marketing or advertising industry. Below, we will discuss examples of AI video and animation production.
The first example is The Dog and the Boy, planned by Netflix, an OTT service. It is known that all backgrounds of this animation were created with AI, and only post-processing was done by humans. Netflix JAPAN said, “It is an experimental measure that will help the animation industry that is short of manpower,” and “All cuts and backgrounds in the 3-minute video are image-generated AI.” “We used technology,” he explained.9)
<Figure 7> The Dog and the Boy
The second example is SNOW. Snow is a Korean mobile application that provides various contents such as camera filters, stickers, and emoticons. It uses facial recognition technology and artificial intelligence technology to recognize the user’s face and apply various filters and emoticons to create more fun and creative content.10) Additionally, a sharing function is provided within the Snow app so that users can share on various social media platforms. Recently, ‘Snow Park’, an AR theme park based on snow, was opened, and various snow-related products and services were also released. Below is a picture showing the AI avatar creation process provided by Snow.
<Figure 8> AI avatar creation process using Snow
3.3 Current Status of AI Convergence Curriculum in Korea
3.3.1 AI content-related department/major operation status
As the domestic AI craze grows stronger, universities everywhere seem to be focusing on nurturing AI talent.11) Below is a table analyzing the operation status of content-related departments at domestic art universities that promote AI education and have AI in the department name or educational goals that include content to cultivate AI convergence talent.
① Kookmin University
<Figure 9> Kookmin University AI Design Department
The above departments place importance on professional knowledge and experience that combines understanding and insight into future technology and the environment, cutting-edge technologies such as artificial intelligence and big data, and design for the future information industry. Therefore, we are developing an artificial intelligence adaptive curriculum such as big data, artificial intelligence, and data visualization, and it is comprised of subjects that enable user-centered design research and practical skills. Lastly, the goal is to foster artificial intelligence adaptive designers by gaining insight into future technologies and environments with the professional knowledge and practical experience required as future designer capabilities in the new convergence era.
② Seoil University
<Figure 10> Seoil University AI Convergence Contents Department
The above departments deal with curricula to train experts with a business mind in the fields of big data, artificial intelligence, and games. This curriculum consists of creative convergence, field-oriented, self-directed, and character-intellectual education, and places importance on education that solves industrial and social problems and creates value. In particular, we aim to cultivate experts with a business mind in the fields of artificial intelligence, big data, and games, which are established as innovative industries for the present rather than technologies for the future society.
③ Nazarene University
<Figure 11> Nazarene University Smart Media Track
In the above departments, you can acquire professional knowledge in the field of multimedia content production, such as video production, editing, sound editing, graphic design, animation, and mobile programming, and based on this, you can develop your planning and production skills in multimedia content. It deals with the curriculum. Through this, we aim to cultivate experts in the field of multimedia content who can lead in a rapidly changing information society. We focus on learning basic multimedia knowledge and multimedia content planning and production skills, and provide major specialization (internship, start-up club, intensive education, Overseas employment program, double major, etc.)
④ Sehan University
<Figure 12> Sehan University AI Content Design Department
The above department is building an industry-academic design curriculum by strengthening the technical and managerial capabilities necessary to efficiently develop design in modern, cutting-edge multimedia environments and new technology spaces. Even if AI education is not advertised in the name of the department, existing content-related departments are conducting research on curriculum composition for the introduction of image generation AI into the curriculum, developing online lecture content in the form of shared lectures, and providing information on existing subject content. Attempts are being made to apply AI as a tool, assigning tasks and conducting research, and the educational goal is to foster designers with the ability to integrate art, technology, and cultural content, which are the essence of design.
<Figure 13> shows the process and results of a class on character design using image generation AI at the Department of Visual Design at Tongmyung University.
<Figure 13> Example of AI Character Assignment from the Department of Visual Design at Tongmyung University12)
The following picture is an example of the current stage in which AI tools can be used in the video content production process. The subjects corresponding to the example items can be nomadized and designed by applying AI to the performance process of each subject.
<Figure 14> Example of Possible use of Generated AI According to the Video Content Production Process
Considering the above possibilities, the subjects in content-related departments are subjects in which Image generation AI can be applied as a tool to the processes of existing subjects, and subjects in which research on continuously evolving and rapidly changing AI technology is designed through self-directed projects. There will be a direction to open new courses that incorporate project-based teaching methods. First, if we try to nomadize subjects that can introduce AI technology as a tool into existing subjects, it is as follows.
<Table 4> Subjects that can Introduce AI Technology Related to Video Content
3.3.2 Suggestions on re-establishing capabilities and introducing curriculum for cultivating AI adaptive content talent
In the future content industry, understanding of new media technologies and content and performing roles as technicians and managers are required, as well as the ability to develop content using new technologies. To this end, content-related departments always need to research and operate curricula to foster talent with the skills and capabilities necessary for the future content industry. Currently, the huge revolution in Image generation AI cannot be resisted, and in order to adapt to this trend and cultivate talented people who can utilize it, the educational program is structured to provide students with more practical experience and abilities by learning theory and practice in an integrated manner and utilizing major specialization strategies. And through this, we should be able to train field-oriented content manpower. In addition, industry and academia should work together to establish an industry-academic education system and, through this, strive to acquire core competencies for efficient content development and role performance. In terms of the curriculum, it should be comprised of subjects that provide creation/production, planning/research, and practical skills by operating an AI adaptive content production curriculum using systems such as big data, artificial intelligence, and data visualization. Finally, in the new convergence era, the new technology and industrial environment should be led with the expertise and practical experience required by future content experts to re-establish the necessary capabilities and design talent
The capabilities required for AI-applied talents using image generation AI are suggested in the following six.
First, it is the fundamental production ability of content, aesthetic sense, and judgment based on formative knowledge. Rather than considering the results of image generation AI as absolute, it should be possible to judge image information generated by AI with theoretical production ability, aesthetic sense, and formative knowledge-based judgment.
Second, communication skills. The ability to express and judge one’s creative ideas, planning, and directing capabilities based on the results is important, not only to use the images derived from image generation AI as they are.
Third, it is the ability to solve real problems. High-level results can be produced only when the subject or problem of a given project can be accurately understood and appropriate prompts can be utilized.
Fourth, creativity and humanities imagination. It is to increase and maintain the ability to lead the tool with the mindset that one cannot replace one’s creativity while creating the image one wants with image generation AI.
Fifth, it is the ability of image generation AI literacy. It is important to analyze and understand the results generated while using AI, continue to verify them, and create new things based on your own judgment.
Sixth, it is self-directed learning ability.
You need the ability to sensibly judge the generated results and learn about your field without relying excessively on image generation AI.
4. Conclusion
This study proposes the application and necessity of AI convergence content-related curriculum, focusing on the innovation of content-related major curriculum in response to the AI transformation era and the spread of AI video content production. Currently, it is urgent to re-establish talent and innovate the curriculum to foster AI-based professionals, which is not only using AI tools, but also has the ability to understand and design the principles of AI software. Compared to the current development of AI technology and the speed of application in industrial sites, educational application and curriculum research are insufficient. Therefore, based on this study, it is expected that it will contribute to fostering experts who effectively utilize AI technology in content-related fields by expanding and utilizing educational methodology exploration, curriculum design, and teaching methods.
References
- Park, S. W. and Lee, H. Y., "Artificial Intelligence Cartoon-like Style Content Research Modification Works", 2023 Korea Information Technology Application Association Spring Conference Paper Collection, 2023.
- Karras, T., Laine, S., and Aila, T., "A stylebased generator architecture for generative adversarial networks",. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 4401-4410.
- Stable Diffusion, namuwiki, accessed 23 July 2023. Available at https://namu.wiki/w/Stable%20Diffusion.