How to Write Prompts for Kling 3.0: Pro and 4K Guide

Why Kling 3.0 needs a different kind of prompt

Kling 3.0 from Kuaishou is built for cinematic motion and physics. It follows physical, concrete action descriptions and strong camera language. It responds to what a camera could actually record: a body shifting its weight, a door swinging on its hinges, a lens pushing in. Abstract mood words like 'epic', 'emotional' or 'stunning' give it little to work with. Verbs and nouns give it a lot.

On Biggsfield you can use Kling 3.0 in two versions. Both support multi-shot generation and native audio, and both share the same prompting rules. Everything in this guide applies to either version. The main differences are resolution and cost.

Kling 3.0 Pro vs Kling 3.0 4K: which one to run

Kling 3.0 Pro renders at 1080p and is the value option. A 5-second clip with sound costs 140 credits (about $1.35 on Creator), 8 seconds costs 224 credits, 10 seconds costs 280 credits and 15 seconds costs 420 credits.

Kling 3.0 4K produces true 4K output with the same multi-shot and native audio support. A 5-second clip with sound costs 350 credits (about $3.37 on Creator), 8 seconds costs 560, 10 seconds costs 700 and 15 seconds costs 1050.

Both models offer durations of 5, 8, 10 and 15 seconds. A sensible workflow is to draft and iterate on Pro until the action, timing and camera are right, then run the final version on 4K when you need the resolution for large screens, cropping or a premium deliverable. The prompt carries over unchanged because the two models read prompts the same way.

Rule 1: Describe concrete physical action

Write the action the way a stunt coordinator or choreographer would. Say who moves, which body parts move, in what direction, how fast and what the movement affects in the scene. 'A woman is sad in the rain' is a mood. 'A woman in a soaked wool coat stops on the curb, lowers her umbrella to her side and lets the rain hit her face, her shoulders dropping' is a performance Kling can render.

Include physical consequences, because Kling is strong on physics. Mention splashes when a foot hits a puddle, dust lifting when a box lands, fabric snapping in the wind, a coffee cup rippling as a train passes. These details make motion feel weighted and real.

Keep each shot to one or two clear actions. Stacking five unrelated movements into a few seconds usually produces rushed or muddled motion. If you need more action, give it more shots.

Rule 2: Use real camera language

Kling 3.0 rewards strong camera direction. For every shot, name the shot size (extreme wide, wide, medium, close-up, extreme close-up), the angle (low angle, eye level, overhead) and the movement (static, slow dolly in, tracking alongside, handheld, orbit, crane up). If you care about the look, add lens character, such as a wide lens with deep focus or a long lens with compressed background.

Tie camera moves to the action. 'The camera tracks alongside her at waist height as she runs' tells Kling how the camera and subject relate. 'Dynamic camera' does not. Biggsfield's AI Video studio also includes more than 20 camera moves, including crash zooms, dolly zooms, FPV and bullet time, which are added to the prompt in each model's own words. This is useful when you know the move you want but not how Kling prefers to hear it.

Rule 3: Structure multi-shot prompts as self-contained shots

Kling 3.0 supports multi-shot generation, so a single run can cut between several shots. The rule from the model guide is specific. When you request shots, give each shot its own self-contained prompt with subject, action and camera, plus a duration. The durations add up to the total clip length.

Self-contained means each shot must make sense on its own. Do not write 'she turns around' in Shot 2 and assume Kling remembers who 'she' is and what she wears. Restate the subject briefly, for example 'the red-haired courier in the yellow rain jacket'. Repeating the key identifying details in each shot helps keep the subject consistent across cuts.

Plan the durations against the lengths available: 5, 8, 10 or 15 seconds. A 15-second clip might be three 5-second shots. A 10-second clip might be 3, 3 and 4 seconds. Give longer durations to shots with more movement and shorter ones to inserts and reactions.

Rule 4: Write dialogue in English

Both Kling 3.0 versions generate native audio, and the native voice only speaks English or Chinese. For English-language content, write every spoken line in English inside quotation marks, say who speaks it and describe the delivery, for example quietly, out of breath or with a dry laugh.

If you need the final video in another language, generate the scene with English dialogue first. Then use Biggs Voice or the Translate & Lip-sync app, which translates speech and matches the lips to the new words in one step.

Native audio covers more than speech, so describe the soundscape in plain terms: rain on a tin roof, distant traffic, footsteps on gravel. Keep spoken lines short. A line that fits naturally in the shot's duration sounds far better than a paragraph squeezed into a few seconds.

Four example prompts

The examples below follow the rules above. The first is a single shot, the second and third are multi-shot sequences and the fourth focuses on dialogue. Each works on Kling 3.0 Pro or Kling 3.0 4K. Choose the duration in the studio to match the total length of the shots.

Single shot, 5 seconds: A barista in a black apron slams a portafilter into the espresso machine, twists it locked and pulls a shot as thick crema streams into a glass cup. Steam hisses from the wand beside her. Close-up at counter height, slow dolly in toward the cup, shallow depth of field, warm morning light from a window on the left. Sound: the machine's hiss, the clack of metal, low cafe chatter.

Multi-shot, 15 seconds total. Shot 1 (5s): Extreme wide shot of a lone cyclist in a red jacket climbing a switchback mountain road at dawn, mist in the valley below, static camera on a long lens. Shot 2 (5s): The cyclist in the red jacket stands on the pedals, legs pumping, gravel spitting from the rear tire; low-angle tracking shot moving alongside at wheel height. Shot 3 (5s): Close-up of the cyclist in the red jacket cresting the summit, sitting back and exhaling hard as the sun breaks over the ridge behind him; camera cranes up and over his shoulder to reveal the view. Sound: wind, chain clicks, heavy breathing.

Multi-shot, 10 seconds total. Shot 1 (3s): Overhead shot of a hand placing a matte black wristwatch on a slab of wet slate, water beading around it, static camera. Shot 2 (3s): Macro close-up of the black wristwatch's second hand sweeping past the twelve, a single water drop sliding down the crystal, slow push in. Shot 3 (4s): Medium shot of a man in a grey suit fastening the black wristwatch on his wrist as he walks out of a glass office lobby; handheld camera follows from behind and swings around to his side. Sound: water drips, a soft mechanical tick, footsteps on stone.

Dialogue, 8 seconds: Medium two-shot in a dim diner booth at night, neon sign buzzing in the window. An older detective in a rumpled trench coat slides a photograph across the table toward a nervous young man in a hoodie, then taps it twice with one finger. The detective says in a low, tired voice: "You were there. Don't tell me you weren't." The young man swallows and glances at the door. Camera slowly pushes in from the end of the table. Sound: buzzing neon, rain against the window, a fork scraping a plate.

Testing and refining your takes

Change one variable per iteration. If the motion is right but the framing is off, rewrite only the camera line. If the action feels floaty, add physical consequences rather than more adjectives. Iterating this way on Kling 3.0 Pro keeps costs predictable before you commit to a 4K render.

Biggsfield's AI Video studio lets you run one idea on up to four models at once and keep the best take. This is a quick way to check whether Kling suits a concept or whether another model handles it better. You can also lock the first and last frame of a shot when you need a clip to start or land on a specific image. If you need the same character or product to appear consistently from reference images, Kling O3 Pro is the Kling variant designed for that, and it supports up to four references.

When your clips are ready, Biggs Cut lets you trim, arrange, add music and voiceover and burn in auto subtitles on a multi-track timeline without leaving the platform.

Questions

More guides