Module 3.2 · Track 3 · AI Filmmaking Workflow
Scene Construction in Freepik
The node-based pre-production workshop. Here you don’t generate isolated images—you build a controlled visual sequence, where every shot has a purpose, narrative logic, and rhythm. Script, character, rough storyboard, and cinematic render, connected through references.
What you will understand
- Why Freepik Spaces is a node workflow — and why this is real pre-production, not random generation.
- The four-step workflow: script → character → rough storyboard → cinematic render.
- How to use references @ to connect the script, character, and previous results across nodes.
- A direction logic behind a six-shot sequence: space → character → action → movement → outcome.
1.Controlled sequence, not a random image
What this step no is: generating random images and hoping they fit together. What it is: building a controlled visual sequence, where every shot has a purpose, narrative logic, and emotional rhythm.1 This is how AI is used in real pre-production and visual development — not as a machine for making pretty images, but as a planning table where the film is conceived before it exists. The Freepik Spaces is the environment where this happens, through a node-based workflow.
A workflow for in the is a visual chain: each box handles a task (writing the script, creating the character, sketching the storyboard, rendering) and passes its result along. Instead of a single prompt trying to do everything, you divide the work into connected stages — and can review, replace, or regenerate each stage without redoing the rest. This structure is what turns “asking for an image” into “directing a scene.”
The four steps in the workflow
The lesson’s pipeline has four steps, and each one is a node (or a pair of nodes). Step 1 — Script: break the story into shots and define the sequence. Step 2 — Character: create the protagonist's visual identity in the image generator. Step 3 — Sketch storyboard (in the assistant): generate rough composition sketches—focus on framing, blocking, and pacing, not detail. Step 4 — Cinematic render (second assistant): translating sketches into realistic, colorful frames, as a previsualization stage. Three honest caveats run through it all: quality depends on clear, precise prompts; you connect the nodes with references @; and the first result won’t be perfect — iteration is normal.2
▸ Going deeper: the complete project for this lesson optional
Optional layer. The instructor provides the entire Space, with all nodes and prompts set up.
The source project includes the node-by-node workflow used to build the structured cinematic storyboard in Freepik — it’s worth opening and exploring the connections between nodes as you read this lesson: Freepik Scene Building (Freepik Space). The logic applies to any scene; Space is just the environment where it was assembled. Freepik is paid, but powerful precisely for node-based pipelines.
2.Script and character: the first two nodes
Everything starts with the script, and the script here is an exercise in breaking things down: take the story and divide it into shots, defining the sequence. It’s not literary prose—it’s a shot list with a purpose (SHOT 1, SHOT 2, ...). This node is the backbone that all the others will reference; if the sequence is clear here, the storyboard and render inherit that clarity. If it’s vague, the vagueness spreads.
The second node is character: create the protagonist’s visual identity in the image generator. As we saw in the pre-production pipeline, this is done under controlled conditions — one consistent character, with a stable face and wardrobe, who can reappear in every shot without drift.3 An ethical warning that the source makes explicit: keep the character generic — do not replicate any identifiable real person.
The script node — a clear shot list, ready for the following nodes to reference:
Stop and predict
Why is it worth spending time making the list of plans e o character clearly defined before touching the storyboard — instead of starting with the sketch and making adjustments as you go?
See one possible answer
Because they are the referenced columns through all the following nodes. A vague script creates a vague storyboard, which creates a vague render — the error propagates and multiplies the cost. Solving the sequence and identity at the source is what maintains consistency across all six shots.
3.The sketched storyboard
The third does not use a assistant to generate rough composition sketches—the focus is framing, blocking, and rhythm, not detail. This is where you work out the scene's progression while it's still cheap: a sketch in black and white, pencil style, showing each shot's readable silhouette without getting lost in micro details. The instructor explicitly asks for a rigid structure — six frames in a 2x3 grid, each with a horizontal cinematic aspect ratio (16:9, with letterboxing), without inventing new shots or changing the order.4
Prompt precision matters because, left to its own devices, the generator tends to create square frames, roughly six of them, or to merge shots. That’s why the rules are emphatic and prioritized: a single image, clean grid, exactly six panels, widescreen content within each panel. The more explicit the layout rule, the less the generator improvises.
The sketch storyboard prompt, with the layout rules that keep the grid under control — ready to paste into the assistant:
▸ Going deeper: why sketch before rendering optional
Optional layer. This is the logic behind spending one step just on rough sketches.
The outline settles the decisions ways to change — framing, blocking, pacing, scale — while they still cost seconds. Black and white and low fidelity are not limitations: they are the point. Without color or texture, the eye judges only the composition, which is exactly what needs to be right before the finishing touches. Rendering first and judging the composition afterward leads to rework: you end up redoing six polished images because the sequence wasn’t working from the sketch stage.
4.The cinematic render
The fourth in — a second assistant — translates the sketches into realistic, colored, rendered frames, like a previsualization stage. Here the goal changes: the result should look a real live-action film still — not an illustration, not concept art, not a stylized render. It's the point where the composition approved in the sketch gains light, color, and materiality.5
The key is in the double reference: the render uses @sketch as a composition reference and @character as the character reference. The critical goal is recreate the storyboard from the sketch as faithfully as possible — preserving composition, camera angles, and spatial relationships — while rendering everything in realistic cinematic quality. Exactly six images are generated, one per shot, each a horizontal widescreen frame (~16:9), with no grid, no contact sheet, and no multiple panels in a single image.
The render prompt, with the double reference @sketch + @character — ready for the second assistant:
It's worth repeating the source's warning, because it's what separates expectation from frustration: the first result no comes out perfect. The assistant may get the composition wrong, change a character detail, or ignore a layout rule. Iterating is part of the craft — regenerate, adjust the reference, and keep the frames that really match the sketch.
5.The sequence’s directing logic
Why this does a six-shot sequence work? Because each shot has a dramatic purpose, and their order forms an arc. Shot 1 (wide): establishes space, isolation, scale. Shot 2 (character): changes to create human emotion and connection. Shot 3 (action close-up): creates tension through the gesture. Shot 4 (inside the car): changes perspective, builds immersion. Shot 5 (movement): the action becomes movement, the energy rises. Shot 6 (wide again): returns to scale, creates closure and continuation.
The narrative arc, in one line, is: space → character → action → movement → outcome. Three principles support the sequence: each shot has a defined purpose; varying shot scale creates rhythm; composition guides attention. Together, they make the scene express — in this example — loneliness, determination, and forward movement. Notice that the wide shot back in the end, it isn’t repetition: it’s what closes the loop while leaving the door open to the next scene.
What you have at the end — and where you go next
By the end of the workflow, you have a shot-by-shot script, a consistent character, a sketched storyboard, and six rendered cinematic frames—a previsualization package of the scene. These frames aren’t the destination: they’re the raw material for the next lesson. In Module 3.3, they gain movement in Seedance. Sum up the module in one sentence: in Freepik, you don’t generate images—you build a sequence, shot by shot, referencing each stage and iterating until the scene has directorial logic.
Stop and predict
Shot 6 repeats the wide framing of shot 1. If you cut and ended the scene on shot 5 (the movement at its peak), what would change in how the sequence feels?
See one possible answer
The scene would lose its closure. Ending at the peak of the movement leaves the viewer suspended, with no breathing room—it can work for an abrupt cut, but sacrifices the sense of a complete cycle. The wide shot on the way back restores the scale, closes the gesture, and, by returning to the transformed starting point, opens the way to the next scene. Repetition with a purpose isn’t redundancy: it’s punctuation.
Before moving on: four quick checks
No grades, no score. Answer from memory, then reveal the answer to compare.
01What’s the difference between “generating images” and what Freepik Spaces proposes?Reveal
Freepik Spaces builds a controlled visual sequence into us, where every shot has a purpose, narrative logic, and rhythm — real pre-production. It’s not about generating random images and hoping they fit together: it’s about dividing the work into connected, controllable stages.
02What references are for @ between the nodes?Reveal
For connect script, character, and previous results. In the render, for example, @sketch gives the composition and @character gives it its identity — that's what keeps it consistent across all six shots.
03Why is the storyboard drawn as a black-and-white sketch, without detail?Reveal
To judge only the composition — framing, blocking, pacing — while changes are still cheap. Color and texture distract from what needs to be right before the finish. Solving the progression in the sketch avoids redoing six polished images later.
04What’s the arc of the example’s six-shot sequence?Reveal
Space → character → action → movement → outcome. Each shot has a job (wide establishes, character connects, close-up builds tension, movement energizes, returning to wide wraps up and continues). Varying the scale creates rhythm; the composition guides attention.