Image source: Generated by Unbounded AI

I was wondering how much impact Deepseek has on the gaming industry.

You see, this tool became very popular before the New Year and was called "national-level scientific and technological achievement" by the producer of "Black Myth: Wukong".

For a period of time, Deepseek itself has become a traffic password and a wealthy person. From concept stocks to skyrocketing, major manufacturers in various fields will be connected to Deepseek, and from small to zero-based use Deepseek to make games...Along this trend, maybe "99% of developers" mentioned by Cai Haoyu are really forced to change careers in advance. .

"Against the Cold" and "Peace Elite" are connected to Deepseek

But is this really the case? After chatting with many game practitioners in different positions, Grape Jun came to a different conclusion.

The following is a self-reported account of these practitioners’ experiences:

01

Planning: Brain violence participants , collaborative efficiency improvement tools, search engines

——Some Shanghai Four Little Dragon Game Level Planning

My colleagues around me made demo When it comes to using your needs, you will first ask Deepseek to write a code, and then give a performance demonstration to your technical colleagues. This operation method will allow technical students to better understand the planning needs.

Using the required content, use Deepseek to directly generate corresponding level slices

The picture material comes from Grape Jun simulation, for demonstration only

In addition, Deepseek is used as a search engine , for example, ask it which games have used a certain gameplay mechanism.

I have also been thinking about how to expand the application of Deepseek, because the great advantage of this tool is that it will give references and reasoning processes, so it doesn’t matter whether the results it gives are accurate or not-I myself It can be seen and a secondary search is performed based on these intermediate results.

This feature is particularly suitable for personal cross-sectional learning and development:

For example, help me sort out the logic of planning documents and check for omissions and fill in the gaps, which is very meaningful for my personal development.

I feel that Deepseek may be similar to the "random" of the past. No one imagined that "random" could produce "Roguelike", and no one could imagine how much potential AI has for gameplay.

——A single-player game copywriting planner

Deepseek last time he was helpful to his work, he should have helped me find information.

In the Internet and deep thinking mode, let it give the names, information, selection reasons, packaging plans, and packaging logic instructions of historical figures suitable for the second packaging of a certain era, and then manually adjust it. Its second filtering and modification is provided as a supplygrade selected materials.

The picture material is from Grape Jun simulation, for demonstration only

But I think Deepseek cannot really join the core R&D workflow for the time being. It will indeed provide some help in the inspiration and packaging level. But the efficiency of finished product conversion is not as high as expected. One is the reason for making up random things, and you need to spend energy cross-verification; the other is that its own model corpus has a distinctive style (such as science fiction), and it is difficult to break into the style you want. At most, some Prompts are given, but this It will lead to very average output results.

In my opinion, it can only take on the function of being a thug at most, and I usually use it to tell fortunes a little more.

——A certain independent game producer

In the planning process, we will communicate with Deepseek about some problems we encounter. It will give some suggestions and directions, especially reasoning. The process of problems can help us check for omissions or provide some new thinking directions:

——A certain Shanghai Four Little Dragon gameplay planning

I also often use Deepseek and other tools as Search engines, or self-censoring tools.

For example, after writing the planning plan, say to it, "You are now playing a test engineer to check it out, and see what edge conditions are there, I didn't expect, and I will question me."

Then it began to be fooled by me to "work".

The picture material is derived from Grape Jun simulation, the game is fictional, for demonstration only

I think the benefits of AI to check for omissions and fill in the gaps are greater than those for projects, because this link is between you and other colleagues When connecting and collaborating, it will be correct, but after using these AI tools, you will become more "careful and reliable" in everyone's eyes.

In addition, I personally have two main aspects of AI tools and applications: creating business tools and building gameplay prototypes.

For the former, although our company has specialized technical middle platform to help plan and develop various tools, it may cost a lot of money from demand, schedule, communication, execution to implementation. Using AI tools such as Deepseek and GPT will save a lot of time and energy.

For example, last time I spent an hour writing a relatively complex business tool that can realize data visualization and verification. Before I didn't have this tool, it might take me a day or two to review the table data exported by these editors, and now it's basically only 5 minutes.

Data verification tools

The situation of building gameplay prototypes is similar, because many manufacturers have project establishment from bottom to top, and we need to review the boss whether what we have made, and in the past, we have done these things. It may require a solution or a PPT, but even if you say it so much and cannot verify it, you still don’t know if it’s fun.

So we need to simply create a prototype of the gameplay, which may not need to be perfect or packaged. It can look like it and achieve the demonstration effect.

At this timeIf you follow the original model and development cycle, it may take one or two weeks and labor costs, and you still need to worry about not passing the review after it is made, and you will be able to revise it back and forth; but with the iterative evolution of AI tools such as GPT, we only need to dismantle it. It is necessary to distinguish what functions are needed for this gameplay prototype, and then talk to the AI ​​tools one by one, optimize and combine them... In the end, it may only take one day to create the gameplay prototype.

When I was doing these things, I was also thinking: Is this considered a "sneck salary"? Because people around me really rarely use AI tools to improve efficiency.

But looking back, even if you really use hand-made code, there will be prediction input tools like Copilot to improve efficiency by 20%-30%. This should be considered another "technical dividend" Now.

Copilot makes predictions based on annotations, the image source is Zhihu @Jayden

02

program: Hu Even making up things must be done as

——Natural planning (technical direction) of a major manufacturer

My current programming work DeepSeek is not used very much because it is too divergent, and I like to remind me to pay attention to some issues that I don't need to pay attention to.

First of all, my work is divided into two parts: one is to find a solution, and the other is to implement the solution.

In the former, I will ask: What methods can be implemented? What I get at this time is basically inspiring answers, and it is still useful; but when I say that I think one of the solutions is good, it will keep moving around this solution and cannot get out.

For example, I asked first: In Unreal Engine's GAS, how do you usually determine whether a Gameplay Ability can be triggered (for example, it can be triggered when the health volume is less than 90%)? It will give me some valuable reference;

The image material is derived from Grape Jun simulation, for demonstration only

My second question is: Suppose there is now a source object and a For target objects, I need to rely on the status of the source object and the target object as the available conditions for Ability (for example, the health volume of both parties is less than 90%). Can I use Gameplay Ability System to make this judgment?

DeepSeek gave me a complete solution: combine attribute listening, conditional check logic and network verification to achieve it.

The picture material is from Grape Jun simulation, for demonstration only

But do you know what the answer I need most for the second question?

is "can't". Ability cannot get Target information before activation. As long as I know this, I can give up this direction and find other ways.

In other words, when I ask a specific question, sometimes I ask "Can", it will think of a way to give me "Can” answer. But if it is really not possible, the way it thinks may turn into making up, or at least it will waste time. Therefore, DeepSeek can only be used in the selection of plans, and it will not be possible after entering a plan. It's out.

But DeepSeek's effect is pretty good in learning, assisted understanding, and story creativity, and it may be used more in the creative side.

03

Art: Prompt Word Optimization Master and Creative Implementation Tool

——A certain game art outsourcing provider Technical Arts (TA)

I mainly use Deepseek in three aspects:

< p>The first is to generate image prompt words, which can accurately understand my creative needs and convert them into professional Prompt. When combined with other AI tools such as Claude and Imagefx, a standard one can be generated in just 5 minutes. Promotional poster. Claude generates poster suggests Deepseek to analyze and optimize and generate accurate Wensheng Picture Prompt Word IMAGEFX uses prompt words to generate pictures manually and adjust details

Secondly, I will use Deepseek to do some creative extensions, such as letting it generate some game worldview and core gameplay architecture for reference.

As well as the scene design and combat details under this framework:

Based on this prompt word, the card concept map generated by Recraft:

The third is to satisfy your own horizontal development and creativity Implement.

I am a complete liberal arts student. However, with the help of AI, I developed my own web pages and mini programs. Even APP has made some gadgets to make the work more efficient. Interestingly, I have my own IT department where I work now, and they are much more professional than me. But after I know that I know how to program AI, they sometimes come to ask me. Some expertise.

For example, I used three days of spare time to successfully develop a decompression applet using Deepseek, Windsurf and Cursor. During the development process, by talking to AI, it can Accurately implement a series of functions I expect. Ultimately, the mini program has the ability to record the mood of workers, give encouragement and reminders to workers, and call Interesting functions such as sucking meditation and encouraging gags.

Of course, I also know that the above three AI workflows may have existed in the GPT era. But the advantage of Deepseek is that it can be used in the Chinese context Fully understanding of my context and needs reduces the cost of repeated communication and the threshold for use. Especially when dealing with some slightly complex tasks, its reasoning ability and networking can provide many ideas expansion.

— —Art design of a major Guangzhou factory

Our company's planning colleagues have used AI to generate renderings more frequently, and then asked art colleagues to go to the renderings.Reach to the touch.

The image material is derived from Grape Jun simulation, for demonstration only

Sometimes the pictures produced by AI are very different from the direction we want, and the proportions of characters are incorrect, so it is more specific. Characters and decoration designs still require us to think about them; but if this AI rendering matches the "feeling" of planning, we can directly imitate them, basically without modifying or refining these manual operations, which is relatively labor-saving.

But I always feel like this is not working properly.

02

Audio: Competitors, Efficiency Improvement Scripts

p>

——Lu Xiaoxu, founder and CEO of Xiaoxu Music

The foreign audio music generation mainly uses two software, Suno and Udio. After the Suno v3.5 and v4 versions were launched this year, it The maturity of AI music creation is already very high.

But it still has two flaws: one is that the sound quality of the generated audio is low, and it is OK to listen to it on a mobile phone, but you can hear the obvious difference by using the console game or using it on a hi-fi speaker. ; Second, Suno cannot accurately control and modify it, so how does the music generated at one time be determined. If you say that the piano in the second is to be changed to a guitar, it cannot be changed.

So Suno has little impact on our gaming audio and music businesses. The demand we receive is still 100% manual production, and AI may be used as a reference in terms of direction at most.

As for Deepseek, last year we have been studying the prompt words for AI music and how to talk to AI tools, but after Deepseek came out, we tried it and the prompt words it wrote were more accurate than what we studied ourselves. All the work in the first half of the year was in vain.

More than this, after AI has entered, many jobs in the audio and music business are no longer needed. For example, game dubbing requires a recording studio. We used to have 4 recording studios in Beijing and equipped with 4 recording engineers, but now that TTS (Text-to-Speech, text to voice) has matured, many manufacturers are not able to record; for example, now The production and singing of many game songs have been replaced by AI, and the workload of singers and CVs will be reduced quickly, and the number of talents will be reduced accordingly.

NetEase Fuxi's AI dubbing has been implemented in the mobile game "Against the Cold" and "Eternal Disaster" mobile game

In this case, I expect our traditional business may It will reduce by 80% within three years. If AI creation is truly mature and commercialized, it will definitely not fall into professional fields or some experts - game companies can use AI to generate any music. Why do you have to pay for someone else? What AI replaces is not labor costs, but a complete set of business processes.

So we are also changing our needs, such as making some content that combines AI audio and video to reduceSubstitutability; for example, try some live performance and interactive performances; at the same time, I am also asking the team for comprehensive abilities, such as the composer having the ability to perform musical instruments, after all, live performances may be more valuable than music production in the future.

Xiaoxu Music combines a variety of AI content short video hits

As for the anxiety problem you talk about, I might be more anxious last year, but this year I’m relatively calm. After all, the general trend of AI is the anxiety of all mankind, not the subject of our family.

——Shanghai Dachang Audio Planning

I think there will be some prospects for mounting and accessing AI tools such as Deepseek in the creation tools.

I used Deepseek to write a script for the audio editing software Reaper. In short, it is to integrate some commonly used plug-ins for software into the additional interface. Click the button to load the corresponding plug-ins, saving The time cost of reviewing the plugin list one by one.

98% of the code of this efficiency improvement script is generated directly by me using Deepseek, and debugging and modifying it through dialogue with AI:

Some problems also arise. After talking to Deepseek to a certain number of rounds, it begins to make up randomly, and the output code becomes completely unavailable. Fortunately, it had already given a relatively complete and basically available code. Based on this code, I manually fine-tuned it based on the Deepseek inference content, and finally realized the implementation of this small script.

In the later Deepseek chaos state, it is not OK to copy the complete code to it

There is no way to use it

05

Operation: Better assists than ChatGPT

——Northern Light Game Operation

I am looking for promotion When creating content such as copywriting and text messages, Deepseek will be used, and it feels that it is much easier to use than ChatGPT at that time. But once things like operation announcements are sent directly to players, millions of people must take a look at them themselves and there is no mistake at all.

Deepseek directly generates Valentine's Day promotional text messages (game information is fictional, for demonstration only)

——Tianmei Game Operation

We have many AI tools within us . But I don’t use it much, so I will use it to check the information, but I just take a look and help. I used it to tell fortunes some time ago... It was difficult to meet the demands when working seriously.

06

Conclusion

From the above Judging from the practical experience of practitioners, Deepseek and other AI tools provide more help to personal work, and are far from bringing some revolutionary changes to the traditional game development process.

This conclusion is also expected. As Lu Xiaoxu said, AI is a tool of equal rights. Its original intention may not be to let professionals use it, but to allow laymen to do professional things without any threshold.

A certain game art outsourcing boss

So, AI has not had that much impact on the gaming industry?

It seems that it is not right. Chen Chao, the CTO of Qingpu Network, believes that everyone's understanding of Deepseek is similar to when other AI tools first came out: they overestimated their short-term effects and underestimated their Long-term role - The impact of AI workflow on the gaming industry is more about penetration than a revolution.

These "penetrations" are like the practical explorations of the above practitioners. They seem to be not serious, but they may become the early accumulation of AI's transformation of the game industry. Cao Xiaowen, founder of Qingpu Network, also said that many manufacturers have used AIGC to replace artificial art assets. They may not have been able to get a regular ROI, but through early accumulation, they have lifted the level of AI in the gaming industry.

"If the human structure of the gaming industry is a pyramid, then AI is a primary neural network, which will be submerged from bottom to top like sea water. The first thing that will be replaced may be some scraps , but with the continuous development of AI and the influence of the structural characteristics of pyramids, the submersion speed of seawater will become faster and faster, even exceed our imagination; this submersion is not just a one-to-one replacement, and it is very likely to make the working structure , model and culture will change together. ”

“When the seawater of AI cannot pass through the middle of the industry, a greater turning point may come.”