From the Script to Scene

In professional filmmaking, every scene begins long before the camera rolls. The director, cinematographer, production designer, and concept artists work together to turn a written idea into a visual plan. In this lesson, we replicate that process with AI tools — and the goal isn’t to “generate images”; it’s to simulate a real film production, from script interpretation to final frames.1 By the end, you’ll see how one creator does the work of an entire pre-production team.

Production always follows a structured process, and that’s what removes chance and replaces it with cinematic intention: script → storyboard → character → outfit → environment → shot list → final frames. Each step feeds into the next. We’ll follow this same sequence in Freepik Spaces, starting by creating a in the text with the scene description.

The question that comes before any image

Before creating any visual, a director always asks one fundamental question: what should the audience feel at this moment? In our example — a samurai in his thirties crosses a long suspension bridge between two mountains, in fog and rain, katana in hand, sensing that something is watching him — the emotions are isolation, tension, and anticipation of the confrontation. These will determine camera angles, lighting, and composition. Every scene can be broken down into four coherent elements: character, o environment, a atmosphere e a narrative tension.

Notice that the emotion defined at the beginning doesn’t get left behind—it runs through the entire pipeline. “Isolation” becomes negative space in the sky; “tension” becomes a narrow bridge; “anticipation” becomes the ready katana and the alert gaze. Defining the feeling first is what weaves every stage into a coherent scene, instead of a collection of beautiful, disconnected images.

The storyboard is the first language of direction. It defines the framing, camera perspective, visual rhythm, and escalation of tension—before any detailed image exists. Even great filmmakers use storyboards to understand how the camera moves, where tension appears, and how a scene unfolds. In our workflow, we use one in assistant to turn the script into a six-frame sketch that shows the moment’s cinematic progression.2

The instruction you give the assistant matters: ask it to act as a professional storyboard artist, analyze the scene and create six panels that visualize the progression of the moment—each panel a different shot, following the cinematic shot scale, while keeping the same location and environment, with a focus on framing, composition, and tension. Then we generate the images for the panels (in the example, with the Nano Banana Pro model). One honest note about the craft: the generator rarely gets everything right on the first try— iterate, regenerate and keep the panels that best follow the desired progression.

This is the storyboard prompt, ready to adapt to your own scene:

In traditional production, this stage belongs to the departments of casting is concept art. In our workflow, the AI takes on that role. The key principle: the character must be created in a neutral studio environment, to isolate the design. The focus is on realistic proportions, a plausible clothing base, natural light, and a neutral pose. This creates a clean reference of the character — which can then be dressed and placed in the environment.3

The character is at the center of the scene, which is why it’s worth insisting on a believable face, natural proportions, and convincing clothing: if the audience believes in the character, they start believing in the scene. Next, we dress this character. In film, this is the work of the costume and props departments; in our pipeline, we generate the clothing and weapons as separate studio reference objects — in the samurai example: the traditional garments, the kasa hat, and the katana, each rendered in studio lighting before being combined with the character.

Why separate everything

Separation isn’t bureaucracy—it’s the default logic of professional visualization pipelines. Building the character, costume, and prop as separate pieces is what guarantees consistency: the same face, the same outfit, and the same katana can appear in all six shots without drifting. Then we connect the character and costume nodes to the image generator, with an explicit instruction to preserve the same character’s identity and proportions.

The wardrobe integration prompt, maintaining consistency:

In professional VFX pipelines, characters and settings are created separately — standard practice in professional visualization. We follow the same principle: we generate the environment as its own piece. In the example, it becomes a dark wooden suspension bridge between cliffs, above a deep canyon, with heavy rain, strong wind, and fog moving through the abyss.4

The environment defines the scale and dramatic tension of the scene, and each element pulls in a different direction. A bridge narrows creates vulnerability — there's nowhere to run. A canyon immense creates cinematic scale — the character looks small against the world. The fog introduces uncertainty — you can't see what's coming. Together, these elements amplify the dramatic impact. Notice how the environment embodies the emotion decided in Section 1: isolation and tension stop being words and become geography.

The environment prompt, derived from the script and storyboard:

Once the scene is clear, we turn the storyboard into a cinematography shot — the shot list. It defines, for each shot: the shot type, the camera position, the camera movement, and the dramatic purpose. At this point, AI acts as both female director and director of photography, helping shape how the audience will experience the scene. A shot list is, in practice, a production document: it determines where tension appears and how the story manifests visually.5

To generate it, we connected to the assistant node everyone the elements already created—the character, environment, storyboard—and ask for a realistic list of six shots based on the scene, referencing each component with the system’s mentions. Then comes the separation of frames: we broke the list into individual images, one shot at a time. Here, again, the honesty of the craft: the generator doesn’t get every shot right on the first try — iterate, regenerate and keep collecting the images that best match the original layout, paying attention to composition and camera design until you have the complete set.

What you have at the end

By the end of the workflow, you have a script, a storyboard, a drawn character, wardrobe and props, an environment, a shot list, and the final frames—in other words, a complete pre-production package a scene. These frames aren’t the destination: they’re the raw material for the next lesson. They’re used for AI video generation, animation, film previsualization, and pitch presentations. The still frame is now ready to come to life.

The shot-list prompt, with AI in the role of director and DoP:

Capture the module in one sentence: one person, with AI and a structured pipeline, does the work of an entire preproduction team — as long as they decide on the emotion first and iterate without rushing. The assignment (in film school format) is precisely this: build your own short scene with this pipeline and experiment with composition and camera design. In the next lesson, we’ll bring these frames to life.