Blog

The Scriptwriting Modus Operandi (Script to Video)

When writing a script intended for Higgsfield, standard screenwriting formats (like Hollywood scripts) must adapt to the way AI models process visual information.

1. Keep the Scene Structure Visual and Action-First

AI does not understand abstract concepts, feelings, or dialogue cues like “he feels sad.” We must write physical, observable actions (e.g., “His shoulders slump as he staring down at a cracked glass”). Every scene must be immediately translatable into physical assets.

2. Frame-by-Frame Scene Breakdown (The Beat Table)

When we write your script, every scene will be broken down into a Beat Table of 1 to 10 visual beats per 15-second clip.

• 1 Storyboard Frame = 1 Beat.
• Density Rule: A standard 10–13 second Higgsfield clip can comfortably fit 4 to 6 beats (visual shots). If a scene has more than 10 shots, we will automatically split it into multiple 15-second clips.

3. Continuous Camera & Environmental Rules

• No “Exit and Re-entry”: If a character exits the frame, the AI considers them gone. We must end the clip or cut to a new beat if a character moves out of sight.
• Camera Movement Lexicon: We will use clear, physical camera commands that Higgsfield handles beautifully: slow dolly in, whip-pan left, low-angle static, tracking shot.
• Atmospherics & Textures: We will add rich, physical descriptions like fog, dust motes floating in light rays, wet asphalt, or warm directional sunlight to anchor the visual style.

What to Enter Into Higgsfield (How the Prompts are Built)

For each 15-second movie scene we generate, the prompt we feed into the Seedance 2.0 model must follow a highly specific, continuous format. There are no line breaks or headers in the final prompt, but it is assembled out of these 6 precise blocks:

[0] Reference Mapping + [1] Style & Mood Preset + [2] Narrative Summary + [3] Starting Composition + [4] Dynamic Description (Beats) + [5] Audio + [6] Negations

The Formula Breakdown:

1. [0] Reference Mapping (Optional but Highly Recommended):
   • If you have a character reference or a sketch, we upload it first and tag it: @Image1 is the character reference — match the character's appearance exactly.
2. [1] Style & Mood Preset:
   • You must choose a visual style first so all clips match. For a cinematic movie, we use: Cinematic photorealistic footage shot on ARRI Alexa 35, anamorphic lens, shallow depth of field, natural film grain, color grade applied uniformly to the entire frame. (Followed by your mood modifier, e.g., Low-key side lighting, deep shadows, slow imperceptible dolly. No subtitles, no text overlay, no captions.)
3. [2] Narrative Summary:
   • One high-level sentence summarizing the clip: A detective sitting in a dimly lit office examines a suspicious folder.
4. [3] Starting Composition:
   • Lock down Frame 1: Opens on a medium shot of a cluttered wooden desk. A cynical 45-year-old detective in a rumpled trench coat sits behind it, holding a manila folder.
5. [4] Dynamic Description (Beats):
   • This is where the script’s action unfolds. Format: [Framing], [camera movement]: [action with physical detail].
   • Example: Tight close-up, static: his eyes narrow as he flips open the folder. Medium shot, slow dolly in: he pulls out a glossy photograph, his hand shaking slightly. Low-angle static: he leans back, staring blankly at the ceiling in shock.
6. [5] Audio Description:
   • Seedance 2.0 generates audio natively in one pass with the video. We must describe the atmosphere and any spoken dialogue: Audio: Low ambient rumble, rain drumming against the window glass, a soft, slow piano chord.
7. [6] Negations (Anti-Slop):
   • Keep the AI from adding junk: No text, no labels, no watermarks, no logos, no UI elements. No color cast, no lens distortion.

The Workflow: How We Build Your Movie

To turn your idea into a movie, we will execute these sequential phases:

1. Phase 1: Concept & Style Selection — We lock down your plot, characters, and visual style (e.g., Cinematic Sci-Fi, Dark Noir, or Pixar 3D Animation).
2. Phase 2: Screenplay Scene Breakdown — I write the script and divide it into sequential, 15-second “Clip Blocks.”
3. Phase 3: The “Hero” Character Generation — To maintain character consistency throughout your movie, we will generate a single high-quality “Anchor Portrait” of your main character(s) first. We register this face as a persistent element in Higgsfield.
4. Phase 4: Video Generation — We batch-submit the clip prompts using the persistent character element.
5. Phase 5: Post-Processing & Stitching — Once the clips are ready, we stitch them together, adding background music or custom transitions.

Let’s begin! What is your movie’s concept, genre, or style? If you have a rough draft or character details already, share them and we will structure the first clip blocks together.