The AI wave is sweeping the game industry: in the next 5~10 years, games may be generated by AI

A year ago, the first people who lost their jobs because of AI appeared in the game industry. Outsourcing teams such as game original painting and translation have begun to encounter cuts, and the development speed and level of AI painting tools have almost comparable to that of human painters with professional backgrounds and junior experience.

A year on, more and more game makers are adapting to the existence of AI as a potential competitor in the capillaries of game production.

Tencent, the largest game company in China, recently came up with the game AI engine GiiNEX, which uses generative AI technologies such as large models to provide 2D images, animations, plots and other capabilities. NetEase used large models and AI algorithms to drive non-player characters and upgrade UGC creation tools in the mobile game "Against the Cold" and "Egg Boy Party".

"What AI can do today is even greater than it was in the last decade. NetEase CEO Ding Lei said in a recent earnings call that NetEase is actively preparing for this round of technology, especially in games, music and other fields. Nvidia CEO Jensen Huang even predicted in an interview at the GTC conference that AI-generated games will appear in 5~10 years.

"From 2022 to 2023, the changes brought about by the 'explosion' of AI tools have been very drastic, and since last year, it has been a deep change, and the iteration of AI tools has accelerated. Lin Rui, art director of Shengqu Games, told the first financial reporter that he has been asked about related topics many times this year, and the public is very curious about AI applications. In fact, more and more experience accumulation and tool enhancement have deepened the depth of AI use, and more and more company teams now have their own dedicated AI tools and large models. He told reporters that ideally, AI can now help complete more than 70% of game art work.

In addition to medium and large teams, there are also small teams that have decided to change the game production process in a more radical way. "There is no outsourced art, all of them are produced through Midjourney, a literary and student diagram tool, and there are only 2 people full-time, one curating and one technology. Wu Yufu, chairman of Firestone Holdings and venture capitalist, told reporters that he incubated a small game team at the end of last year and is currently developing AI games.

From drawing to retouching

According to industry insiders, the "competitor" that game artists have to face is the equivalent of AI with 3~5 years of work experience after graduation.

"Generally speaking, a person from a beginner to a professional practitioner has a long learning time, and AI tools have saved us at least three or five years. Lin Rui told reporters that in Shengqu, AI has entered the original painting, general design, UI, font generation and project resource management in game development.

The last wave of AI "subverting" the game artist profession was in late 2022 and early 2023, when the emergence of AI mapping tools caused some outsourcing companies to lose orders, and order prices plummeted. Within game companies, changes are also taking place, and they are now gradually manifesting themselves in efficiency improvements. "According to the previous process, it may take 1~2 weeks to draw a Meixuan, and there are differences between different projects and different periods, but AI can be completed in a very short time. In the same cycle, AI brings more choices and possibilities. Lin Rui told reporters.

Feng Bin, an expert in game technology and art, also has obvious perceptions. "We used to have a meeting once a week, and this week we might have only done part of the role, and we're going to continue to talk about it next week. Now I have two meetings a week, and I can have several characters to discuss in both meetings. Feng Bin said that his project uses the open source and free Stable Diffusion, and the project team uses a suitable manuscript to train the large model, and finally the AI drawing is not much different or even better than the human drawing, and the artist modifies the AI-generated picture on the basis of it. "The artist 'complained' that he had changed from a painter to a retoucher. He said.

Shengqu is also vigorously promoting the use of AI tools, and its art production mainly uses self-developed and free AI tools, including secondary development on the basis of Stable Diffusion. During ChinaJoy in August last year, Shen Fengliang, vice president of Shengqu Games and head of the art center, said that 2D graphics account for about 20% to 30% of the game art, and AI tools can roughly replace half of the workload, about 10% to 15%, and some 2D games can replace a larger proportion, which may account for 30% to 50%.

Lin Rui told reporters that the penetration of AI tools is still strengthening, and when most of the process and cost of game projects are concentrated in art, the more art resources need to be completed, the greater the role of AI tools. After making preliminary preparations, such as training a style model based on the art style of the project, AI can complete more than 70% of the art work in this ideal situation. However, style model training is no longer a difficult task, and a stable workflow has been formed.

From the explosion of AI mapping tools to the gradual penetration in the past year, the changes in game production have been gradual.

Lin Rui told reporters that at the beginning, the game production staff still used AI on the basis of the original work, and now they are moving towards a more pure AI creation method, using AI to replace some of the original work links. In the past year, including internal research and development and external open-source tools, the models and tools available for game art have increased, and the model has been trained more subdivided.

Wu's team directly used the paid closed-source Midjourney, and decided to use it after the AI tool iterated to a certain level last year. "Before Midjourney, we didn't find a tool that met the requirements of game production, but after Midjourney came out, AI became a viable path. Wu Yufu said that this is the reason why his team does not have full-time art positions today.

"For small and medium-sized game companies, AI lowers the barrier to entry, improving efficiency, shortening time, and reducing costs. You don't need an artist, as long as you have a programmer and a planner, you can bring up a team. Wu Yufu said that the cost reduction brought by AI tools is also calculable, and the game company he founded used a team of 100 people to spend 3 years to make an RPG online game "Water Margin Q Biography", including 50 artists, 10 original paintings, 20 modeling, 20 3D actions, and the cost is 10,000 ~ 15,000 yuan per person per month. Up to now, the Wensheng diagram AI tool has been able to replace the original painting part, and the original line drawing and color drawing positions are no longer needed.

The cost of using AI tools is not too high. "For example, I have a team that does light casual battles, and it can take a week or two to give it to an outsourcing company to generate three maps. If there is still a battle lighting effect in the game, the art production will take about 3 months, and the cost will be about 80,000~100,000 yuan. With AI, the cost is basically negligible. If you pay about 20 yuan with Midjourney, you can get 50~60 round-trip maps, and at least 20 maps can be generated, with an average of 1 yuan per map. Wu Yufu said.

Who else will replace?

Leading companies in the industry are already responding to this new round of technological change.

In response to analysts' questions on Tencent's recent earnings call, Chief Strategy Officer James Michelle said: "In terms of the use of AI in games, and in many cases, the line between hype and reality depends on how far one is willing to look forward, and how far we are willing to look." Such as AI-driven non-player characters, AI-accelerated graphical content generation, graphical asset generation, etc., all of which will benefit greatly from AI deployment in the coming years. "

Not long ago, at the NVIDIA GTC conference, Perfect World announced that its terminal game "Zhuxian World" will be connected to NVIDIA's Audio2Face technology, which can use generative AI to convert audio into animation. In addition, Nvidia CEO Jensen Huang predicts that the era of AI-created games could arrive in the next 10 years, and early attempts could be seen in five years, with GPU-powered AI tools used to generate computer graphics.

In fact, this isn't the first time the gaming industry has experienced a disruption of jobs with new technologies and tools. Some practitioners told reporters that about 20 years ago, 70% of artists in a R&D team were artists, and with the upgrading of the tool chain, the proportion of planning, technology and art has been basically the same.

But the impact of this round of AI may be more far-reaching than ever.

After AI eliminated a number of outsourced art companies last year, AI is now putting pressure on the in-house artists of game companies. "The pressure on outsourcing companies is estimated to be relatively high, and the demand for art manpower by game companies has also decreased, and large manufacturers may cut some art that is not so good in technology. Just retouching the pictures can be done by experienced artists, and the senior artists are basically not affected, but the novices are greatly affected. Feng Bin said.

"AI tools are suitable for reducing costs and increasing efficiency in the production of a single resource, a single direction, and a single type of resource. In the R&D team, whether it is art, planning or technology, there will be fewer and fewer entry-level positions to complete a single standard and a lot of repetitive labor. Lin Rui told reporters that for positions like compound such as resource integration, including project art framework construction, quality control, and roles that need to make breakthroughs in existing art styles, AI cannot penetrate at present.

However, Lin Rui also believes that under the iteration of tools, the impact of AI on game production will not be limited to the current 2D art, and after the new tools enter the existing work pipeline, they will be transmitted upstream and downstream, and eventually form a new work pipeline. The future of video, audio, animation, and 3D production will gradually change. The entry of AI will have some positive impact on the industry, or create more valuable and higher-level jobs.

What other changes will AI bring to game production in the foreseeable future, and what else can human employees do to cope with the impact of new technologies?

"Thankfully, games are a human, creative industry, and humans need to play a role. AI lacks the expression of 'humanity', and humans are good at putting the finishing touch. AI may one day replace humans, but understanding the characteristics and competitive strategies of competitors may delay the time for humans to be replaced. Lin Rui said.

In addition to AI mapping tools, video generation tools such as Sora or world simulators have become the new favorites in the field of artificial intelligence this year, and the industry is also evaluating the impact on game production. A number of industry insiders told reporters that it is unlikely to replace the game engine, on the one hand, the game engine is a physical engine, and Sora does not show enough understanding of the real world, on the other hand, if the real-time online game uses AI to generate plots and pictures in real time, the network latency and the computing power behind it are extremely demanding, and the cost of computing power may also be extremely high, which is not realistic at present.

However, Sora's ability to produce coherent images is enough to change the art process once again, according to Wu. "If Sora is open to use, the generated video will be made up of many photos that can be handed over to the technical hands, and some games where the entire art team can basically skip over, and games like "Water Margin" will only leave 10% of the art for acceptance and adjustment. He said that given the high efficiency and low cost of AI, "a game that originally cost 100,000 yuan to buy a batch of art materials, if Sora can replace it, I would rather spend 10,000 yuan to buy Sora services, saving 90% of the money and time." "

AI may not be able to be used indefinitely in games in the short term. Lin Rui mentioned the problem of computing power, now a MOBA game in a game, may have used seven or eight AI, the AI chat system in the game has a large consumption of AI computing power, if the cost of computing power continues to increase, whether it can achieve good enough business performance, the result is not certain. After AI lowers the threshold for game production, creativity will become more important, and given that the game market has not changed, and the arrival of AI has not necessarily significantly changed the market seat, the logic of how the game team should survive and live well has not changed.

Wu Yufu also believes that after AI penetrates and lowers the threshold of "making a game", the future game will become an era of "talent is development", and creativity can win. In addition, large companies may rebuild their organizational structures or even reduce some personnel expenses to achieve higher profits with the help of AI. Larger companies may also develop separately from smaller teams and make relatively heavier products.

(Intern Hao Zizhu also contributed to this article)