How AI Art Is Like Directing

Understanding the role of an AI creative director.

On a large film set, the director does not touch camera. They don’t act. In many cases, they don’t even write their own script. The director “prompts” other artists to carry out their vision. AI art works the same way, with a creative human directing AI. While the director is understood to be the “author” of the film, since AI is a newer artform some still mistakenly credit the tools rather than the creative who wields them.

When photography first began as an artistic medium, it received many of the same criticisms as AI. Photographers just “push a button” and the camera “does all the work for them.” Over time, we’ve come to understand that the choice of where to place the camera, how to frame and capture the subject matter, and what subject are photographed are all artistic choices that reflect the intention of the human artist more than the tool of a camera. In time, AI art will be understood the same way.

Detractors of AI art claim that the lack of control makes AI tools different from other artforms, yet anyone who has worked on a film set knows that the directors also do not have total control. The other artists they collaborate with bring their own ideas that are not the director’s. The director’s creative choice is whether or not to incorporate those ideas into the film. When the director includes a contribution of their collaborators, those choices become the director’s artistic choice as well.

For example, the classic film Taxi Driver (1976) contains a scene where Robert De Niro’s character says “You talkin' to me?” while imagining a confrontation alone in his room. The line is on the American Film Institutes 100 Best Movie Quotes at #10 and considered one of the most iconic moments of the film. Who is the “author” of this line? The film is written by Paul Schrader, but that line was improvised by actor Robert De Niro. Yet despite neither writing nor saying the line, director Martin Scorsese is consider the “author” of the film to the point where the movie is sometimes referred to as Martin Scorsese’s Taxi Driver.

While Schrader and De Niro have significant creative roles in the film, the reason Scorsese receives credit is that he made the choice to include those creative contributions. He created the environment where De Niro could improvise that line and chose to include that improvisation in the final film. A lesser director might have told De Niro to “stick to the script,” edited it out of the movie, or never created an environment where the actor could improvise in the first place.

Likewise, AI might have unplanned moments where it creates genius that the human creative prompting it did not plan. Even in these cases, the human creative deserves credit for the choice to create a scenario where AI could create that genius and the human choice to include that result in the final product. Sorting through the takes of another creative to find the best ones and arranging them in the final artistic composition is its own artform, known as “editing.”

The idea that art requires total control is a myth. Many artists create by allowing an artistic process to lead to their final product rather than a predetermined target. Most improv and documentary falls into this category, yet both mediums still have a director whose artistic vision receives authorship credit. Even productions with a predetermined script are often improved by “happy accidents,” where unplanned mistakes that deviate from the original vision manage to improve the final product.

In the original planned vision for Jaws (1975), the shark worked. Part of what separates director Steven Spielberg from lesser artists was his ability to incorporate unplanned events like the animatronic shark in his shark movie not working, even if that meant deviating from his original vision. Now, many consider Jaws a better film because of how little the shark is seen. The limitations and mistakes of AI could be framed in a similar way if used artistically by an equally competent artist.

Like directing, AI art often involves a “workflow” rather than a single prompt. The director has an artistic vision, but they typically hire department heads to handle each part of the film who also direct - director of photography, art director, etc. A director might know what the look and feel of a film needs to accomplish, but they will allow an art director to handle specifics. The art director might then hire other artists to further handle specific aspects of the art direction, like a costume designer who knows how to create period accurate costumes. If the production requires enough costumes, the costume designer might hire a seamstresses who sews their designs.

If we were to frame this collaboration in AI art terms, there is a “workflow” where the director prompts a vision to their art director, who translates that into a specific prompt for the costume designer who does a design that the seamstress sews. In this case, the seamstress isn’t the “real” artist, while everyone else is just taking credit for their work. They are part of a larger artistic process.

Many AI art workflows follow a similar pattern, where a creative turns their intention into a script in ChatGPT that becomes a series of prompts in MidjourneyAI which create still images that are used by Runwayml to create a moving video. Complaining that the AI artist doesn’t create each individual element by hand is a bit like complaining that the director doesn’t sew their own costumes. Like a film production, there is a workflow of creative processes that carry out the creators vision using artforms they do not directly practice themselves.

AI turns every artform into directing. You don’t have to draw to create art. You don’t have to play an instrument to make music. You don’t have to program to build an app. Practitioners of these artforms sometimes complain that AI creators aren’t “real” artists in the same way they are. They are correct, in the sense that directors are not "real” actors like the leads of their films. They are practicing a different artform. It’s true that Martin Scorsese can’t do what Robert De Niro does as well as him. However, Scorsese is also practicing a unique artistic craft that De Niro cannot do at his level.

We can appreciate different artforms without demeaning or deriding other artists. Comparison is the thief of joy, and framing AI art in opposition to traditional art misses all the ways that artists can collaborate with AI and other artists who use it. Good directors acknowledge the value of the other artists they work with. We should also acknowledge the artistic value of direction, even if it comes from an artist who does not directly practice the artform they are directing.

Studios seeking to replace their creatives with AI will soon discover that directing is its own artform and studio heads are not as competent at directing as actual directors. Much of the backlash against AI art is due to bad directing. While a director could use the first draft of a script, the first rehearsal take, or the first edit of the film, most know better. Non-creatives don’t. Those who’ve never done creative work often see the outputs of AI art programs as better than anything they could create on their own and publish the results without taste or revision. Because AI can create quickly, they flood the internet with their first drafts. The result is often derided as “slop.”

Since competent creatives using AI tools in artistic work and non-creatives who don’t care about the output of their “content” both fall under the label “AI art,” the public often sees bad direction as intrinsic to the tools rather than a choice by those using them. Imagine the confusion the public would have about directing if both Christopher Nolan and someone uploading their camera roll to YouTube were referred to as “directors.” Although both are technically creating moving pictures, there is an artistic distinction between them. The same artistic distinctions apply to different forms and uses of AI in art, but AI is so new as a creative tool that most of the public is not yet aware of these distinctions.

Most AI art is bad, but 90% of everything is bad. However, AI moves so fast that users can make the bad art faster than possible in previous artforms and automation is itself part of the artform. AI artists would benefit from focusing as more on the “art” part of AI art than the “AI” part, since general audiences care more about the emotional impact of what they see than how it was made.

If you want to know why AI hasn’t produced a classic yet, consider how new the artform is. The general public has only had access to AI art tools for about two years. Motion film was invented in 1888 and the first public screening of films was in 1895. Name a classic film from 1897. It’s taken over 100 years for film to become the artistic medium it is today, and we are still discovering new ways to use it. Expect AI art to continue to evolve as well.

You don’t have to understand how any art is created to appreciate it. Most of the public doesn’t understand how movies are made. They just enjoy watching them. Likewise, you don’t need to understand AI to know whether or not a piece of AI art moves you or doesn’t. However, misconceptions or false believes can hinder the reception of art. As someone who uses AI, it’s my hope that this framing of AI creator as director allows others to understand, appreciate, and better use these tools.

The next time you see something created with AI, consider the human choice to prompt, edit, and share that output instead of all the other creative choices that person could have made. AI art is human art.