08 = Newsletter Individual #7

Issue #7

Jun 18, 2026

8 min read

You’re Prompting AI Video Wrong. Here’s the Fix.

My model-by-model breakdown of what Veo, Runway, and Kling actually respond to, and why the same prompt that works in one tool produces garbage in the other two.

Subscribe for Free

Share this Issue

What you’ll get in 3 minutes

TLDR

In this issue

8 min read

Big Idea

Diagnose

Match

Prompt

Every major AI video model in 2026 accepts a text prompt. That is where the similarity ends. Veo 3.1 was built to think like a system: give it structured scene data, reference images, and explicit audio specifications and it follows instructions with 87% adherence across complex multi-element scenes. Runway Gen-4.5 was built to think like a shot designer: give it a strong first frame, describe the camera behavior, and iterate fast around motion. Kling 3.0 was built to think like a director: describe multi-shot choreography, human movement, and the audio-visual logic of the sequence. Writing the same prompt across all three is not a workflow. It is three failed generations in a row.

“A prompt is not a universal language. It is a conversation with a specific model. The ones who get consistent results are the ones who learned the difference.”

I spent four months running the same briefs through all three models with identical prompts, then model-specific prompts, then comparing the outputs. The delta is significant. Not just in output quality, but in the number of failed generations before a usable clip. Model-specific prompting reduced my per-clip generation failures by roughly half. The prompting vocabulary is learnable in an afternoon. The time it saves compounds across every production week after that.

The 4-Step Workflow

Step 1

Understand What Each Model Is Listening For

Three models. Three completely different prompt philosophies.

Veo 3.1 responds to structured scene data. It wants to know who is in the frame, what the environment contains, what sounds should accompany the visual, and what the camera is doing, all in explicit, layered detail. It reads prompts like a production brief. Runway Gen-4.5 responds to shot design language. It wants a starting frame, a camera behavior, and a motion direction. Feed it a reference image and describe what the camera does next. Kling 3.0 responds to directorial choreography. It wants to know the sequence of actions, the multi-shot logic, and the audio-visual relationship across the clip. Think of it as briefing an assistant director, not a cinematographer.

Generic prompt → Model-specific language

Step 2

Match the Prompt Structure to the Job

The model you pick should follow from what the shot needs, not from habit.

Use Veo 3.1 when the shot requires synchronized dialogue, complex environmental sound, or multi-element scene accuracy. Its 87% prompt adherence rate on complex scenes is the highest of the three. Use Runway Gen-4.5 when you have a strong reference image and the primary variable is camera motion and character consistency across clips. Use Kling 3.0 when the shot involves realistic human motion, multi-shot storytelling, or a sequence where audio and visual need to be choreographed together. If the shot involves fast-moving people, Kling’s human motion rendering is the most reliable.

Familiar tool → Right tool for the shot

Step 3

Write the Prompt in the Model's Native Language

Veo wants a production brief. Runway wants a shot card. Kling wants a director’s note.

For Veo 3.1: open with subject and environment, layer in audio specifications, specify camera movement, close with format constraints. For Runway Gen-4.5: upload your reference image first, then write the camera behavior only. Do not re-describe the character. The image handles that. For Kling 3.0: write the action sequence in order, specify whether you want Smart or Custom multi-shot mode, name the shot types and transitions, add audio last. Each model has a different entry point. Finding it is the skill.

One prompt style → Three prompt structures

Step 4

Test Across Models Before Committing Credits

Higgsfield lets you run all three from one dashboard.

Before committing full-quality generation credits to a shot, use Higgsfield’s multi-model dashboard to run the same prompt through Veo 3.1, Runway Gen-4.5, and Kling 3.0 in draft quality. Higgsfield hosts all three under one subscription starting at $17.40 a month on the annual plan and allows side-by-side comparison without switching platforms or managing separate accounts. The comparison pass costs a fraction of a full-quality generation and often reveals which model interprets the brief most accurately before any real credits are spent.

Blind commitment → Informed model selection

Tool of the Week

featured

Higgsfield

Multi-model AI video platform with cinematic camera controls

The feature that changed how I approach model selection is Higgsfield’s side-by-side generation mode. I write the brief once, adapt it to each model’s prompting style using the Higgsfield Assist prompt copilot, and run all three in draft quality before deciding where to spend final generation credits. Cinema Studio 3.5 also gives real directorial camera controls, dollies, trucks, tilts, and orbits, specified as actual cinematography terms rather than descriptive language. For creators who work across multiple models, this is the most efficient single-platform option in the current market.

Try it yourself

Prompt Recipe

Copy & Use

SCENE: A woman in a navy blazer sits at a glass desk in a quiet office.

She picks up a phone, listens briefly, and sets it down.

Morning light through tall windows. No dialogue.

— VEO 3.1 VERSION —

You are generating a cinematic AI video clip in Google Veo 3.1.

Scene: A professional woman in a navy blazer sits at a glass-topped desk in a modern office. Floor-to-ceiling windows to her left. Morning light. She reaches for a black desk phone, brings it to her ear, listens for 3 seconds with a neutral expression, then sets it down carefully.

Camera: Static medium shot, chest height, very slight push-in over 8 seconds.

Audio: Office ambient noise — faint HVAC hum, distant street sound, phone receiver click on pickup and set-down. No dialogue. No music.

Format: 8 seconds. 1080p. Photorealistic. One continuous shot.

Constraints: No camera shake. No background characters. No artificial sound effects.

— RUNWAY GEN-4.5 VERSION —

[Upload reference image of the woman at the desk as anchor]

Camera behaviour: Slow, deliberate push-in from medium to medium-close over 8 seconds. Subject picks up phone from desk surface, holds it to ear, sets it back down. Motion is controlled and deliberate. Camera continues slow push throughout.

Lighting: Warm morning window light from screen left. No artificial fill.

Duration: 8 seconds.

Constraints: Do not change subject appearance from reference image. No secondary characters. No fast motion.

— KLING 3.0 VERSION —

Shot 1 (3s): Wide establishing shot. Woman enters frame from left, walks to glass desk, sits down. Camera static.

Shot 2 (3s): Medium shot. She reaches across desk, picks up phone, brings it to ear. Slow push-in.

Shot 3 (2s): Close-up on her face — listening, neutral, sets phone down. Cut.

Audio: Office ambient only. Phone click on pickup. No dialogue. No music.

Character: Navy blazer, professional appearance, consistent across all three shots.

Mode: Custom Storyboard. Transitions: hard cut between shots.

Pro Tips:

Mini Case Study

Real Results

Same prompt copied across Veo, Runway, and Kling. Inconsistent results, wasted credits, no clear reason why each generation failed.

After: Model-Specific Prompt Structure

Brief adapted to each model’s native language. Correct tool selected before generation. Credit waste cut by more than half.

Download

FREE

Model-Specific Prompt Cheat Sheet

Ready-to-use prompt structures for Veo 3.1, Runway Gen-4.5, and Kling 3.0 by shot type.

What's included:

Free for newsletter subscribers. By downloading, you agree to our terms of use.

Vani Aggarwal

Filmmaker & Head of Content Strategy exploring the edge where Al meets
story.

Get Al & filmmaking hacks weekly.

Join 12,000+ creators getting actionable Al workflows, tool reviews,
and creative techniques delivered every Wednesday.

No spam, ever. Unsubscribe with one click. Read our privacy policy.

Share & Discuss

Spread the word

Filmmaker & Head of Content Strategy
exploring the edge where Al meets
story.

Start a conversation

Filmmaker & Head of Content Strategy
exploring the edge where Al meets
story.

Reply by email

Related Issues

Explore more AI video workflows and creative techniques from our archive

Issue #3

The Character Consistency Fix Every AI Filmmaker Needs

Runway Gen-4.5's image anchor method in practice. The setup, the results, and where it still breaks.

Issue #4

How I Build a Full Short Film Stack for Under $25 a Month

Three tools, one routing framework. How to send the right job to the right model every time.

Issue #6

Kling 3.0's Multi-Shot Mode Is the Storyboard Tool Nobody Told You About

Six connected shots in one pass. How to pre-visualize a full scene before production begins.

Issue #17

Designing with intention

Learn how to bring purpose into every creative process you start.

Ask Your Questions

Frequently Asked Questions

How to enroll for a Course?

All courses are available at vaniaggarwal.com/courses. Browse the full catalogue and enroll directly on each course page.

Can I get the recordings of my previous lectures?

The newsletter is a standalone publication, not a course. For recorded lecture access, visit your enrolled course dashboard on the courses page.

Who would be the instructor for enrolled course?

All courses are taught by Vani Aggarwal, filmmaker and AI strategy consultant with 16+ years directing across six countries and hands-on testing across 50+ AI production tools.

What kind of placement support will be given post completion of program?

Course graduates receive access to the community, resource library, and direct pathways to Vani’s consulting and coaching programmes for continued support.

GET IN TOUCh

Let’s Create Something –
Extraordinary

Whether you’re a filmmaker rebuilding your generation workflow, a content team cutting credit waste, or a creative director looking for sharper AI production strategy, this is where the conversation starts.