For indie game developers, AI may be more useful than imagined.
With the explosion of AIGC products such as ChatGPT and Midjourney, the game industry is quietly changing. When we talked to a small team of game producers half a year ago, he mentioned that one of the things he was most worried about was that he couldn't recruit suitable artists, the stall was too small to pay too high a salary, and he couldn't learn the program and couldn't art, which affected the development progress overall.
"Bright Memory Infinity" producer Zeng Xiancheng uses AI-generated game protagonist "Shuya"
And now, when we talk about this topic again, he said that he did not have this trouble at all, because soon after the concept of AI painting came out, they tried to use it, and although the AI-generated work still has flaws in terms of details, it is completely sufficient for their puzzle game that is not based on graphics.
In fact, such cases are not uncommon in the current gaming industry. While the head manufacturers are trying to lay out the source and establish a large model platform, and the middle and waist manufacturers are constantly verifying the feasibility of AIGC tool applications internally, smaller and medium-sized and even individual developers with lighter volumes have actually applied AI to game development.
We also spoke with a number of indie game producers about how they are using AI to aid game development.
Increase efficiency by more than 50%, and treat AI as a 24-hour online personal trainer
Aftershock is the producer of the "Strange RPG" game, he resigned from the "Heart Source Studio" under Station B last April and began to try to make his own games, and in October 2022, "Strange RPG" produced by him alone was officially released to Steam. After the official launch of the game, because of a large number of "funny equipment" in the game, it attracted the attention of many players for a while, and caused a wave of second-creation frenzy at station B.
Affected by this, "Strange RPG" successfully exceeded 100,000 sales in just two months after its launch, becoming a dark horse in domestic independent games last year, receiving 94% praise on Steam, and the producer's aftershocks were also concerned by many practitioners.
Today's Aftershock, in addition to updating Strange RPG, is developing a new action game, and in his development process, the AI mainly plays a role similar to that of a "personal trainer". Aftershock told us that the AIGC product we are currently using is ChatGPT, but it does not really integrate ChatGPT into game development, but uses it as an auxiliary tool.
"Mainly through ChatGPT to solve some confusion, such as what are the advantages of two different attack judgment settings. And ask it to help write some simple code." Aftershock mentioned that in many unfamiliar areas, it will be much better with AI, which is equivalent to a personal trainer who can provide 24-hour online guidance, and the use of ChatGPT has improved his development efficiency by more than 50%.
Aftershocks treat ChatGPT as a "personal trainer"
In fact, there are many solo game producers like Aftershock who simply use AIGC-related products to assist game development, such as "Swire Fire" producer She Yue said that he has recently tried to use AI to assist game development, mainly used in line creation, illustration and icon painting production, etc., overall use is relatively rudimentary, but will still improve a lot of efficiency.
"The advantage of AI is to improve efficiency, and the current is to make scraps on the design, such as the plot has to be done by yourself, but the plot illustration, dialogue, these rich details can be first drafted by AI and then optimized by yourself." In addition, She Yue also mentioned that a typical problem in the current use of AI to assist game development is that the higher the requirements, the more difficult it is to control. The time spent on tuning and training can often be directly solved.
"AI has helped me save tens of thousands of yuan in outsourcing costs"
Of course, if the AI is really trained, it will no longer be a simple answer to questions for the development of the game, and it may even save a large amount of development costs for developers.
Tsang Xiancheng (Feiyan Islands), the producer of "Bright Memory Infinite", mentioned that he used Stable diffusion two months ago to generate some character graphics as a reference for the design draft of the game's pre-development. Recently, he used GPT-4's Q&A to allow AI to save labor costs on code scripts.
"Not long ago, I was trying to implement a server communication function, control the opening and closing of the game server, and create my own database to store player data, as well as Https hot update, which is the part of the code that GTP-4 helped me complete, saving about tens of thousands of yuan in outsourcing costs, which is a very good help for independent developers."
"Shuya", the protagonist of "Bright Memory Infinite" generated by AI
It is worth mentioning that in addition to Zeng Xiancheng, there is also a developer who applies AI to game development, he is Wei Xinyu, the producer of "Mountain and Sea Traveler".
In the previous [Dialogue Producer] column, we reported that Wei Xinyu had obtained a doctorate and was engaged in artificial intelligence related work before entering the game industry, so now he uses AI to assist game development, which is quite a bit of "returning to the industry".
Wei Xinyu said that the main tools they currently use are Chat GPT and Stable Diffusion, and the main application scenarios focus on many aspects, such as sending an API request to OpenAi, accessing ChatGPT in the game to allow NPCs to extract the environment information in the game, and then automatically generating dialogue and writing poetry based on the environment information; Use Chat GPT to help plan color/stage performances, scene dialogue similar to magicians, help plan data collection and references, and use Stable Diffusion to simply generate avatars.
Wei Xinyu uses stable diffusion+lora-generated villain-style drawings
For Wei Xinyu, the biggest feeling that using AI to assist game development brings him is not only to make the NPCs in the game more vivid, improve the development efficiency, but even the planning level in the team has been correspondingly improved, and he can quickly write the character settings and dialogue in the field he does not understand.
There is still some way to go to fully apply AI to game development
Improving efficiency is how many game developers feel after using AI to assist game development, but in addition, some developers have mentioned the current limitations of AI and relatively disadvantages.
For example, in addition to the "difficult to control" problem mentioned by She Yue, Zeng Xiancheng also mentioned that although AI in the design of game original art, there are some design schemes that are not particularly high, and the speed of drawing and the efficiency of repeated modification are much faster than manual, but there will also be certain copyright risks.
"However, this problem can also be trained by establishing its own material library, which may require the studio to draw a batch of images as the basic model for training in the early stage, and then when the studio enters the streamlined production stage, use these self-trained models to generate corresponding images, but the energy and cost required in the early stage are no less than the direct use of manual design."
When it comes to using GPT-4 to aid game development, Tsang believes that it has a higher advantage than image generation, but may also spend more time on finishing.
"For some junior program developers, by providing detailed text descriptions, let AI accurately generate the corresponding functional code, speed up the efficiency of writing code, but the disadvantages are also more obvious, if you need to implement a very advanced logic or very complex functions, you need to answer multiple times to let GPT-4 understand the meaning of the developer, for huge projects, AI will not consider the structural management of the project standardization, in the finishing may spend more time than manual labor."
In summary, the disadvantages and disadvantages mentioned by these developers are actually a manifestation of the application of AI to game development is still in the initial exploration stage, which is also a stage that more developers are currently going through.
For example, Yang Yang, producer of "Pascal Contract", Chen Hongqu, producer of "Falling Leaf City", Zhang Shuyang, producer of "Deep Fire" and many other developers have said that they / their team have tried to use AI to assist game development, but due to various reasons such as insufficient maturity of tools, they have not deeply integrated AI and game development for the time being, but their emphasis on AI is the same.
Under the prevalence of AIGC, the game industry is ushering in a quiet and dramatic change, which is both a challenge and an opportunity for the majority of game developers, and how to seize the opportunity of the times to rise to the top is a question worthy of deep consideration by every practitioner.