Can AI drawing assist game creation, and what impact will it have on the game industry in the future?

Thank you for your invitation~ There have been many discussions about AI drawing recently, which may be due to the “Space Opera” created by AI won the Grand Prize in the Colorado State Fair Art Competition, and the “Crow” created by AI for the best short film at the Cannes Short Film Festival get.

Radical people believe that AI will definitely replace human artists, which means that the most sacred and inviolable field of human beings is close to falling. Others believe that AI is just another tool, and its trans-epochal significance will not exceed the invention of the camera, and the camera Obviously, it does not dissolve the value of the artist’s existence, but instead enhances the artist’s creativity.

However, these topics about the meaning of art’s existence are too grand and distant to have any guiding significance for real life and work. In practice, I have seen that some artists have used AI to generate atmospheres, textures, etc., saving the tedious work of detailing – this is the application from the perspective of basic tools.

In the current domestic industry, AI rarely enters the vision of project decision makers (or creators who have the final say) – such as film directors, game producers, fashion managers, etc. I guess that the “bosses” are probably not as interested and capable of receiving new things as they used to be. They may tell their assistants to take a general look, but generally speaking, “bosses” are often arrogant.

However, the art style and quality of a project largely depends on the project decision makers, not the project art executors. We have seen many great artists with amazing personal work, but working on mediocre projects, who are unable to break through the limitations of upper-level decision makers and can only contribute their proficient foundational skills (which may be 30% of their abilities). % less than). Taking the game industry as an example, unless the artist himself becomes a high-level partner of the project, this status quo is difficult to change in the current Chinese game industry – and this prerequisite is not easy.

So what is the real value of AI art at this stage? In my opinion, it is likely to mean more to the art demand initiator than to the art execution party.

– AI can help decision-makers (or Party A) become better decision-makers (Party A), at least for games.

Before the project is established, AI can help decision makers to clarify their own vision and creative intentions, such as subject matter, style, category, etc. The producer can completely hand over the ideas that he came up with by “slap his head” to the AI, add whatever keywords he wants, and then see if the generated result is what he really wants. Before a fully mature direction emerges, the producer can test it repeatedly, because what he needs at this time is not precision, but “feel” – when the word is uttered, it is often the nightmare of his art creators. But AI can obviously work tirelessly to satisfy any unreasonable request.

From another perspective, AI may also bring more adventurous spirit to those conservative producers when setting up projects. The producers of most domestic commercial projects will directly give up the control of the project style in the early stage (this is the most cherished power by creators, but they are often given up directly), and directly select a successful product to carry out. “refer to”.

In projects, AI allows producers to express their needs as clearly as possible when connecting with artists. In the process of work, the biggest cost and time loss often come not from the art itself (art work overtime is meaningless), but from the producer’s inability to clearly express his needs, so that a lot of it is left for the art to guess. There are elusive blanks, and when these parts are submitted, the producer feels “not right”, which leads to a large number of repetitions and even rework of the art. At the same time, the producer is often moved by his strict “craftsman spirit”.

Traditional requirements documents, sketches, photos, or photo p-maps can provide some general reference directions. But it is far less than the AI mapping that describes the atmosphere, lighting, style and texture in detail, plus a layer of self-screening. If you can add p-pictures and document descriptions on this basis, you can communicate Misunderstandings are minimized.

If in the first two steps, the demand side has made a good preparation with AI, put forward a clear vision, and unified the cognition of “what are we going to do”. Then I believe that when the project is in progress, the passion for art creation will also be stimulated. This is where AI comes into its own: revealing as many possibilities as possible with complete unity behind them. The flaws of AI will also be magnified – the details are complicated but lack reasonable choices, the atmosphere is faithfully reproduced but lacks the appeal that humans can empathize with – and this is exactly what human artists are good at.

When discussing with the company’s art colleagues, we came up with the statement that “AI is more like water in an ink painting”. In traditional Chinese landscape painting, painters often use large splashes of ink to quickly fill the picture, and the general shape of the landscape will emerge from the flow of ink, the smudge of rice paper, and the penetration between brushstrokes. At this point, the picture is actually composed of the artist’s guide (“prompt”) combined with the physical properties of the water (“AI”). Next, the artist will use dry brushstrokes to continue to refine the large shape that has been shaped by water – the key here is that the artist does not know how the water will flow in advance, and he cannot be like oil painting. Precise control of the picture can only be roughly controlled, but the trend of water formation in turn will inspire the artist, making him grasp the contingency and determine it with relatively clear brushstrokes.

Party A clarifies the direction of the project, puts forward clear requirements, and also inspires art, just like water in ink painting. This “rich possibility within clear boundaries” should be a benign way of creation.

To test this way of working, I spent a short time with two art colleagues creating a series of illustrations for an imaginary new IP, many of which were not what we wanted to make in the first place. It was AI that generated some kind of inspiration after random generation, and then used a clearer p-picture to determine the image and atmosphere.

This set of graphs is not purely generated by AI. Many graphs combine the results generated by multiple AIs, and some structures have been manually redrawn. I wrote a full backstory, and generated and screened all the images, which two colleagues processed and collaged into narratively meaningful and interesting settings.

These pictures have nothing to do with the “Shadow Blade: The Convict” and “Shadow Blade Zero” that we are actually studying, but the aesthetics and styles contained in it are in the same line, and sometimes even more unrestrained and radical, which is why , AI assistance will definitely enter our formal project process in the future.

Hope AI can bring more possibilities!