How to Create Film Storyboards in Veo 3 – A Step-by-Step Guide

Veo 3 is the latest version of Google’s video generator, and it can do something competitors are still learning – create coherent shot sequences without losing context between clips. You no longer need to edit twenty separate generations in Premiere and pray that your main character looks the same in every shot. Veo 3 remembers what it generated a moment ago and maintains visual consistency throughout an entire scene. In this tutorial I’ll show you how to configure the tool, which prompts work best, and how to control the camera without guesswork.

Google made Veo 3 available in AI Studio in late 2024, but most creators still use it like a basic generator – one prompt, one clip, done. In doing so they miss about 70% of the tool’s capabilities. Veo 3 is not Runway or Kling, which excel at single shots. It is a tool for building entire sequences where each clip flows naturally into the next. Filmmakers in the Google Creators Program tested Veo 3 for four months and developed concrete methodologies. It’s time to share them.

What you need to get started

You access Veo 3 through Google AI Studio – the main platform for experimenting with Google models. You need a Google account, ideally Workspace or a personal account with AI tools access enabled. Veo 3 is not yet publicly available the way Gemini is in the browser, so if you don’t see a video generation option in AI Studio, join the waitlist via the form on Google Labs. It usually takes a week, two at most.

Once you have access, you need two more things: a clear vision of what you want to create and a willingness to iterate. Veo 3 will not produce a perfect storyboard on the first try, but the third or fourth attempt usually hits the mark. Set aside two hours to learn the interface and run your first tests. Google provides a limit of 50 generations per day for basic users and 200 for Workspace Premium – enough to learn the basics and build your first full sequence.

Pro tip ✅

Before you start generating in Veo 3, write out your shot list in a plain text editor. Each shot is a separate line: “close-up of face”, “medium shot from the waist”, “wide shot – full figure in the background”. Veo 3 works best when you know exactly what you expect from each clip.

Setting up your first project

In AI Studio click “Create New” and select Veo 3 from the list of available models. The interface looks similar to other generators – a text field for the prompt, parameter sliders, and a preview on the right. The key difference is the “Add to sequence” button below the generation window. That is how you build a series of connected shots instead of generating random clips.

Start by setting the basic parameters for the whole project. Set the resolution to 4K if you plan a high-quality final export, or 1080p if this is just a working storyboard. Veo 3 in 4K generates clips up to 8 seconds long; in 1080p up to a minute – but for storyboards, shots of 3-5 seconds work best. Short, specific, with one clear movement or camera change. Longer clips lose dynamism and Veo 3 starts hallucinating additional elements.

Set the visual style at the beginning and stick with it throughout the whole sequence. Veo 3 has predefined styles: Cinematic, Documentary, Animated, Photorealistic. For fictional storyboards, Cinematic works best – it adds depth of field, natural lighting, and a light color grade. Documentary suits interview or reportage material. Animated is still in beta; ignore it for now.

Warning ⚠️

Do not change the visual style in the middle of a sequence. Veo 3 tries to maintain consistency, but jumping from Cinematic to Photorealistic between shots looks like a camera change mid-scene. Stick with one style from start to finish.

Your first prompt – how to open a sequence

Good storyboards begin with an establishing shot – a wide shot that shows where we are, who the main character is, and what time of day it is. Veo 3 needs this information in the first prompt because it will reference it in subsequent shots. Don’t write a novel, but do provide context. Instead of “a woman walks down a street,” write “a young woman in a red coat walks down an empty street at dusk, silhouettes of buildings in the background, rain on the asphalt.”

A prompt has three elements: main subject, action, environment – in that order. Veo 3 understands English best, so write prompts in English, but don’t overcomplicate the syntax. Simple sentences, concrete nouns, avoid metaphors. The model doesn’t know what “a nostalgic mood” means, but it knows what “warm golden hour lighting” or “soft focus background” means.

young woman in red coat walking down empty street at dusk, city buildings silhouette in background, wet asphalt reflecting streetlights, cinematic wide shot, 4K

That is the basic establishing prompt. Generate it, wait 20-30 seconds (Veo 3 in 4K likes to think), and you get a clip. If the woman looks right, the background fits, and the movement is natural – click “Add to sequence”. If something is off – edit the prompt and generate again. There is no point building a sequence on a flawed first shot.

Pro tip ✅

Add “consistent character” at the end of your prompt if you plan a series of shots with the same character. Veo 3 will then record distinctive features – hairstyle, clothing, facial features – and try to preserve them in subsequent generations. It’s not perfect, but it improves consistency by roughly 60%.

Camera motion control – how to direct the frame

Veo 3 lets you control the camera through prompt parameters or interface sliders. Sliders are faster for testing, but prompts give more precise control. Filmmakers from Google Creators recommend a mix – basic movement in the prompt, fine-tuning with sliders. The most important parameters are camera movement and motion intensity.

Camera movement has five modes: static, pan, tilt, dolly, orbit. Static means the camera stays still while the subject moves. Pan is a horizontal left-right camera movement. Tilt is a vertical up-down movement. Dolly is a push in or pull out from the subject. Orbit is a circular movement around the character. Each mode changes the dramatic weight of the shot – static gives calm, dolly builds tension, orbit reveals the space around a character.

close-up of young woman's face in red coat, rain drops on skin, camera slowly dollying in, shallow depth of field, emotional expression, 4K cinematic

This is the second clip in the sequence – after the establishing wide shot we move into a close-up. Dolly in builds intimacy and focuses on the character’s emotions. Veo 3 automatically remembers that the character wears a red coat and has a rain-wet face, so you don’t need to repeat that. But “rain drops on skin” is a detail that reinforces continuity – the first clip had rain, and so does the second.

Motion intensity is a slider from 1 to 10. Low values (1-3) produce subtle movement; high values (7-10) deliver action-film energy. For dramatic storyboards stay between 3 and 5. For commercials or music videos you can push to 7-8, but be careful – above 8, Veo 3 sometimes overdoes it and produces motion blur that makes the frame hard to read.

Pro tip ✅

Changing camera movement between shots is the simplest way to build dynamic rhythm. A sequence of static – pan – dolly – static creates a natural rhythm that a viewer’s eye follows easily. Don’t mix orbit with dolly in two consecutive clips – that’s too much movement at once.

Building the sequence – how to connect shots

You now have two clips in the sequence: an establishing wide shot and a face close-up. Time for the third – a medium shot showing what the character is doing. Veo 3 works best when consecutive shots logically continue the action. If in the close-up the woman looks to the left, in the medium shot she should walk in that direction or look at something on the left side of the frame.

medium shot of young woman in red coat opening old wooden door, hand on brass doorknob, camera static, warm light spilling from doorway, 4K cinematic

The third shot introduces a new element – the door and the light from inside. This is a natural transition point where the sequence can branch. Either we follow the character inside (next clip: interior shot) or we stay outside and watch her disappear through the door (exterior wide shot). Veo 3 does not plan the dramaturgy for you, but if you give it a logical chain of events it will maintain visual consistency.

A key rule: every new clip should inherit at least one element from the previous one – character, location, lighting, or time of day. If you jump drastically – from a rainy street to a sunny park – Veo 3 will interpret this as two separate scenes and stop maintaining character consistency. You can use this intentionally for parallel editing, but in that case signal clearly in your prompts that this is a different location or time.

Warning ⚠️

Veo 3 has a sequence memory limit – around 8-10 shots before it starts forgetting details from the first clips. If you are building a longer storyboard, divide it into segments of 6-8 shots, export each segment, then start a new project. You can join them in editing later, and each segment will be internally consistent.

Controlling lighting and mood

Veo 3 understands cinematographic lighting terms better than its competitors. You can write “golden hour”, “high key lighting”, “chiaroscuro”, “neon glow” and the model knows exactly what you mean. This changes how you create mood – instead of describing emotions (“a sad scene”), you describe filmmaking tools (“low key lighting, cool blue tones, shadows on face”).

interior shot of young woman in red coat entering dimly lit room, single table lamp casting warm pool of light, hard shadows on walls, camera pan following her movement, 4K cinematic

The fourth shot – we move inside. A single table lamp is a specific light source that Veo 3 will place in the frame and use to illuminate the scene. Hard shadows build drama. The camera pan follows the character across the room – a natural movement that connects the action from the previous shot (opening the door) with the introduction of the new location.

Color palette matters too. If the whole sequence has cool tones (blues, grays), don’t suddenly throw in warm orange midway – unless it is dramatically justified, for example the character stepping from a cold rainy street into a warm interior. Veo 3 picks up on such contrasts and will use them as a storytelling accent.

close-up of woman's hands taking off wet coat, water dripping on wooden floor, shallow focus on droplets, warm ambient light, 4K cinematic detail shot

The fifth shot – a detail shot of the hands and dripping water. This type of shot builds immersion and gives the viewer a moment to breathe between bigger actions. Veo 3 handles these inserts very well because it doesn’t need to track the whole character – just a detail, a movement, a texture.

Pro tip ✅

Add one detail shot every 4-5 shots – hands on a door handle, raindrops on glass, feet on stairs. These are breathing moments in the edit that give the sequence rhythm and help Veo 3 reset its character tracking if it starts losing facial details.

Advanced techniques – match cuts and transitions

A match cut is a transition between shots where a movement or shape in one frame connects with a similar element in the next. Veo 3 can simulate this if you give it precise instructions. Example: the last clip ends on a close-up of a round lamp; the next begins on a close-up of a round wall clock. The viewer’s eye follows the circular shape and moves smoothly between scenes.

extreme close-up of round table lamp, warm glowing bulb filling frame, camera static, 4K cinematic

extreme close-up of round wall clock showing 11:47 PM, second hand ticking, similar warm lighting, camera static, 4K cinematic

Two prompts, two separate shots, but the composition and lighting are deliberately similar. In the edit this will look like one smooth transition. Veo 3 does not produce automatic transitions like fades or dissolves – you need to add those in post. But it can create visual continuity that makes the cut between shots nearly invisible.

Another technique is leading lines – lines that guide the viewer’s eye through the frame. If in one shot the character walks along a corridor running from the lower left to the upper right of the frame, the next shot should continue that direction. Veo 3 remembers this if you write “continuing movement from previous shot” or “matching screen direction” in the prompt.

Pro tip ✅

Screen direction is fundamental film grammar. If a character in shot A looks to the right, in shot B (the reaction of another person) they should look to the left. Veo 3 will not enforce this automatically – you have to encode it in the prompts: “looking screen left” vs “looking screen right”.

Export and finalizing the sequence

You have a finished sequence of 6-8 shots in Veo 3. Each clip can be exported individually as an MP4 in 4K or 1080p. Google AI Studio also offers the option to export the whole sequence as a zip file with numbered files – convenient if you plan to import directly into Premiere, DaVinci, or Final Cut. The files are already in order, so it’s just a drag and drop onto the timeline.

Veo 3 does not add audio automatically – every clip is picture only. You need to add the soundtrack in post. For storyboards this is usually a plus, since you plan dialogue, music, and effects separately anyway. But if you need a quick preview with an audio placeholder, use AI Studio to generate ambient sound through Gemini – the same interface, the Audio tab.

Most filmmakers use the Veo 3 export as a base for further work in traditional editing. They trim shot lengths, add transitions, color correction, and titles. Veo 3 is a prototyping tool, not a final product. But as a prototype it is good enough to show a client, pitch an idea to a producer, or test whether a sequence works at all before investing in a real production.

Warning ⚠️

Veo 3 generates a Google AI watermark in the bottom right corner of every clip. It can be removed in Premiere via crop or Content-Aware Fill, but if you plan commercial use of the material, check the license in Google AI Studio. Veo 3 is currently in an experimental phase and not all subscription plans allow monetization of generated clips.

Real-world examples – what filmmakers do with Veo 3

A documentary filmmaker in Toronto used Veo 3 to storyboard a 15-minute film about climate change. Instead of drawing traditional storyboards – which he couldn’t do himself and would have cost USD 2,000 if commissioned from a studio – he generated 80 shots in Veo 3 over three evenings. He showed the result to a funding committee, received a grant, and shot the real film. The Veo 3 storyboard now sits in the archive as preproduction material.

An advertising agency in London uses Veo 3 for pitches to clients who have trouble visualizing ideas. Instead of moodboards and animatics, they show finished 4K sequences. The client sees exactly what the spot will look like before the agency spends a single penny on real production. The rate of accepted pitches rose by 40% since they switched to Veo 3.

An independent director in Berlin tests scene blocking in Veo 3 before going on set. She generates different variants of the same sequence – once with a static camera, once with a dolly, once with an orbit – and then on set she knows exactly which setup works best, saving the crew’s time and the budget spent on unnecessary shots. For her, Veo 3 is a virtual location scout and rehearsal space in one.

What Veo 3 still cannot do

Veo 3 struggles with fast dialogue between characters. If you generate a conversation scene, the characters’ lips move randomly and are not in sync with any audio. Google is working on lip sync for Veo 4, but for now avoid close-ups of speaking characters – or frame the shot so faces are not visible: over-the-shoulder shots, profiles, wide shots.

Long shots over 10 seconds lose movement consistency. Veo 3 generates the first 5 seconds well, then starts to improvise. A character who was walking straight suddenly turns sideways. A camera doing a pan stops halfway through. Keep shots short and let the edit stitch them together.

On-screen text – subtitles, signs, lettering – is still a lottery. Veo 3 can generate a “neon sign” but you cannot control what it says. Sometimes you get “OPEN”, sometimes gibberish. If you need specific text, add it in post-production as a graphic. Do not trust the model to read the prompt “neon sign saying HOTEL” literally.

Pro tip ✅

If Veo 3 generates something strange that breaks the sequence – for example a sudden third hand, a classic AI glitch – don’t try to fix it with more generations. Accept that the shot didn’t work, step back one shot in the sequence, and try a different camera angle or a different prompt. It is often easier to go around a problematic frame than to fight the model’s hallucinations.

What comes next – how Veo 3 is changing preproduction

Veo 3 will not replace real production, but it changes the point at which you commit to real production. In the past, a storyboard meant sketches and an animatic meant a slideshow with music. Now you can show a sequence that looks 70% like the final product. This shifts the decision-making moment – you see what works before you spend the budget.

For small productions and independent creators this is a democratization of access. You don’t need USD 50,000 for a test shoot to find out whether an idea makes sense. Veo 3 costs about as much as a Netflix subscription per month and gives an unlimited number of generations within reasonable limits. That levels the playing field between major studios and solo directors with a laptop.

For large productions it is an iteration tool. You can test ten versions of the same scene before deciding which one to shoot. The Director of Photography sees what you are planning and can prepare technically. The location scout knows what kind of places to look for. Everyone on set shares a common vision instead of making assumptions.

Google has announced that Veo 4 will have better character tracking, longer shots, and ControlNet-style tools for more precise composition control. But Veo 3 is already good enough to change the way filmmakers plan projects – and this is not a future two years away. It is happening today, in AI Studio, free for the first 50 generations per day.

Promptyze

Promptyze covers generative AI in plain English — hands-on reviews, tutorials and daily news, fact-checked and hype-free.

See Full Bio