How to Write Long-Video Prompts: 10–15s Multi-Shot Scripts | NadouPro Tutorial

Key takeaways

• Video prompts longer than ten seconds are not piles of keywords; they are shot scripts written in chronological order.

• The five-part framework: scene setup, subject focus, event progression, realistic texture, and motion design. Write them in order so the model understands more accurately.

• For multi-shot sequences, describe the environment before the action, and list the number of shots, shot sizes, and durations one by one so the pacing stays controllable.

• Character consistency depends on locking with reference assets: use multi-angle character images or a short video reference so the model remembers the character’s face and clothing.

• Audio-visual synchronization models can generate dialogue and ambient sound at the same time. When writing the script, clearly mark the speaker, lines, and tone.

Why long videos need “script thinking”

Early video models could only create three- to five-second clips, so narrative was out of reach. A new generation of models supports generating continuous footage of ten to fifteen seconds in a single pass, making it possible to fit a small story with a beginning, development, turn, and ending. Models integrated into NadouPro, such as Kling 3.0 Omni 4K and Seedance 2.5, support longer-duration generation with audio-visual synchronization, and with 3D Director they can also export multi-shot content up to ninety seconds long.

As the capability improves, the writing method has to change with it. Prompts longer than ten seconds must work like a route map: first explain the environment to establish the space, then define the subject and actions in chronological order. This structure ensures the model stays focused on the story you want instead of improvising freely.

How to plan a multi-shot sequence

For complex scenes, you do not need to edit manually. The idea is to write the narrative intent clearly so the generation stage can directly produce a complete scene. There are two planning methods:

Automatic shot breakdown: Write the full plot into one prompt, and the model decides the cut points based on the actions and scenes in the description—when to cut from a wide view to a close-up, when to use a reverse shot, and so on. This is suitable for quick visualization: check the result first, then refine it.

Shot-by-shot customization: NadouPro’s Storyboard supports splitting a script into shots with one click. Each shot can have its own content and duration, and the model follows the shots strictly instead of making its own decisions. This is suitable for narratives that need clear pacing and a complete structure. Take a typical scene as an example:

• Shot 1: exterior wide shot of a European villa, three seconds

• Shot 2: close-up of a woman shaking a glass of juice, four seconds

• Shot 3: medium shot of a man responding, three seconds

Each segment has its own prompt, with the shot size and character actions defined one by one. The split shots are arranged in a column on the canvas, generated one by one, and then assembled in Clip, keeping the pacing controllable throughout.

The five-part script framework

When writing a ten- or fifteen-second script, move away from the keyword-cloud approach and use a modular five-part framework, feeding information to the model in order:

Scene setup: Establish the space and light first. Location, time, atmosphere—for example, “a quiet rooftop at night, city lights in the distance.” The scene description can include sensory details, such as a cool breeze and the distant sound of traffic; these details also guide sound generation.

Subject focus: Identify the main character with a specific description, such as “a woman wearing a blue striped shirt,” rather than a vague label. Reuse the same descriptive phrase throughout the same script to keep the character stable.

Event progression: Describe how the event develops, breaking actions into continuous steps that correspond to the flow of time. A ten-second script should read like a shot list, describing every beat of the scene in order. Writing the script chronologically is the most reliable way to get predictable results.

Realistic texture: State the image-quality requirements, such as “cinematic texture, 4K textures.”

Motion design: Describe how the camera moves—slow dolly in, shot/reverse-shot cutting, orbit tracking.

An example combination:

• Subject focus: Character A and Character B

• Event progression: a conversation about a secret

• Realistic texture: cinematic texture, 4K textures

• Motion design: slow dolly in plus shot/reverse-shot

• Scene setup: outdoor terrace of a villa

How to maintain character consistency

The biggest challenge in long videos is character drift—the audience can immediately notice when the protagonist’s face changes between shots. The solution is to build a “character profile” for the model:

Subject binding: Extract the character’s facial structure, hairstyle, and clothing texture from reference assets and anchor them in the generation workflow. Even when the camera makes a large orbit or a fast push-pull move, the character remains recognizable. In commercial projects, a brand mascot or actor must keep the same face in every scene, so this step is essential.

Two ways to build the profile:

• Multi-angle images: Upload up to four character images from different angles, including front, side, and back views, so the model does not have to guess. NadouPro’s Multi-Grid capability can generate a Character Body 3-View and a Character Expression 9-Grid in one pass, which can be used directly as profile-building assets.

• Video reference: Upload a three- to ten-second character reference video. In addition to appearance, the model can extract movement habits and voice characteristics.

After the profile is built, keep it active through every cut in the entire multi-shot sequence, and the character’s identity will not be lost.

How to script dialogue scenes

Audio-visual synchronization models build sound synthesis directly into the video generation process, so separate lip sync is no longer needed. Dialogue, ambient sound, and on-screen action are synchronized: the system understands the semantic relationship between image and sound, generating the sound of clinking glasses when it sees glasses clink and aligning lip movement when it sees someone speaking.

Requirements for writing dialogue scripts:

• Clearly label the speaker, and place each line next to the character who is speaking.

• Use straightforward text to mark the language, tone, delivery style, and sound effects.

• In scenes with multiple speakers, make it especially clear who is speaking.

• Write the action before the line: for example, the character slams the table first, then starts questioning. This order helps the model understand the physical context of the dialogue.

• Add micro-actions such as breathing and blinking to make the character feel much more alive.

NadouPro’s Audio node provides Voice Dubbing and Voice Conversion capabilities. The generated dialogue voice expresses emotion naturally, and together with the video’s audio-visual synchronization capability, the sound and picture can be generated in one pass.

Camera movement glossary: shots serve the narrative

Cinematic camera language can turn ordinary clips into professional storytelling. Use the right terms in your prompts and you can act like a digital director:

• Dolly in: Builds intimacy or highlights a moment of realization—“the camera slowly dollies in.”

• Low-angle side shot: Shows details and emphasizes a character’s sense of power—“low-angle side close-up.”

• High-angle wide shot: Shows the scale of the environment and creates an epic feeling—“high-angle extreme wide shot.”

• First-person perspective: Creates immersion—“a first-person shot from the rider’s perspective.”

• Orbit shot: Reveals the environment while keeping focus on the subject’s face—“slow orbit camera movement.”

• Pan reveal: Starts from a small detail, then pans out to show a fuller environment.

NadouPro’s photography parameter presets include more than three hundred combinations. Focal Length, Depth of Field, and film style can all be specified directly. When more precise local control is needed, use Motion Control to define motion paths for specified elements, keeping the structure of static objects unchanged while characters move as usual.

Script best-practice checklist

• Choose high-quality mode for commercial projects so textures and motion physics feel more realistic.

• Use the five-part framework to organize prompts and keep the information hierarchy clear.

• Establish the space and lighting before the action begins.

• Use a character profile to lock character traits and prevent identity drift when switching shots.

• Add negative descriptions to filter common issues: distorted limbs and low image quality.

• For dialogue scenes, clearly write the speaker, lines, language, tone, and sound effects.

• Test different motion intensities to find a balance between dramatic tension and image stability.

Frequently asked questions

Q: How many seconds can an AI video be at most?

Most tools generate five to ten seconds in a single pass. The new-generation models integrated into NadouPro support fifteen seconds in a single pass. For longer narratives, use a multi-shot workflow: split shots with Storyboard, generate each segment, and assemble them in Clip, or use 3D Director for multi-camera arrangement, with exports up to ninety seconds.

Q: What is the best way to write long-video prompts?

Organize them chronologically: describe the environment first, then define character actions; specify concrete camera movements such as dolly in and low angle; and use clear nouns and verbs to help the model understand the plot accurately.

Q: How do you maintain character consistency across multiple shots?

Use reference assets to build a character profile: upload multi-angle images or a short video reference. After the model extracts facial structure and clothing textures, keep them active throughout the entire sequence so they are not lost with each shot change.

Q: Can AI generate synchronized dialogue and sound effects?

Yes. Audio-visual synchronization models generate image and sound at the same time. Dialogue is automatically aligned with lip movement, and ambient sound matches the scene. When writing the script, mark the speaker and tone for each line, and multi-person scenes will remain organized.

Q: How do you control camera angles in an AI video?

There are two ways: write camera movement terms directly in the prompt, such as dolly in, pan, and orbit, or use Storyboard to specify the shot size and duration for each shot individually. 3D Director also supports visual multi-camera timeline arrangement for more precise scheduling.

To explore more features, visit NadouPro: NadouPro

Contains AI-generated information that may be incorrect. Please verify independently.