Most people blame the model when their AI video looks fake. Is that fair? Not really. In most cases the PROMPT is the real problem.
Veo 3.1 is a very capable video model. It understands camera language, lighting, motion, and even sound. But it can only work with what you give it. A lazy prompt gives a lazy clip. A clear prompt gives you something that looks close to real footage.
This guide shows you how to write better prompts, step by step. You will get a simple formula, real examples, a few tables, and a list of mistakes to avoid. If you want to test the ideas as you read, you can open the Veo video generator in another tab and try each example.
What Is Veo 3.1 and Why Prompts Matter So Much
Veo 3.1 is Google’s AI video model. It turns text, and in some workflows images, into short video clips with matching audio. That audio part is important. The model can create dialogue, background noise, and sound effects together with the visuals, so your prompt should describe sound too.
Here is the thing many beginners miss. Veo does not read your mind. It reads your WORDS. If you write “a woman walking in a city,” the model has to guess everything else: her age, the weather, the time of day, the camera angle, the mood. Every guess is a chance for the result to look generic.
So what is the fix? Remove the guessing. Tell the model what you want to see, hear, and feel.
The Basic Prompt Formula for Realistic Video
A strong Veo 3.1 prompt usually has seven parts. You do not need all seven every time, but the more you include, the more control you get.
| Part | What It Means | Example |
|---|---|---|
| Subject | Who or what is in the frame | A middle aged fisherman with a gray beard |
| Action | What the subject is doing | Pulling a wet rope onto the boat deck |
| Setting | Where the scene happens | A small wooden boat on a calm lake at dawn |
| Camera | Shot type and movement | Slow push in, medium close up |
| Lighting | Light source and mood | Soft golden sunrise, light mist on water |
| Style | Overall look and feel | Natural documentary footage, shallow depth of field |
| Audio | Dialogue, ambience, effects | Water lapping, rope creaking, distant birds |
Put them together in one flowing paragraph. You can write them in any order, but starting with the subject and action usually works best.
Here is a simple example:
A middle aged fisherman with a gray beard pulls a wet rope onto the deck of a small wooden boat on a calm lake at dawn. Slow push in, medium close up. Soft golden sunrise light with thin mist over the water. Natural documentary style, shallow depth of field. Sound of water lapping, rope creaking, and distant birds.
Short, clear, and visual. That is the goal.
Step 1: Start With a Specific Subject
Vague subjects create vague videos. Compare these two.
- Weak: A man in a kitchen.
- Better: A tired chef in his fifties, wearing a stained white apron, wiping sweat from his forehead in a busy restaurant kitchen.
The second version gives Veo age, clothing, emotion, and place. Notice that you are not writing a story. You are describing what a camera would actually capture.
A few tips for subjects:
- Mention age range, clothing, and one or two physical details
- Keep the number of characters small, one or two is the safest
- Describe emotion through visible behavior, not abstract words
That last point matters. Instead of “he feels sad,” write “he stares at the floor and slowly lets out a breath.” Veo can show a behavior. It struggles to show a thought.
Step 2: Describe One Clear Action
Veo 3.1 clips are short, so one scene and one main action is the right size. People often try to fit a whole movie into one prompt. What happens then? The video looks rushed, or the model ignores half of it.
Good action prompts use simple verbs and clear motion.
- “She lifts the cup, blows on the steam, and takes a slow sip.”
- “The dog runs across the wet grass and shakes water off its fur.”
- “He opens the car door and steps out into the rain.”
Avoid stacking five actions in a row. If you need a longer story, break it into several clips and join them in an editor.
Step 3: Build the Setting With Real Details
Realism comes from small details. A “park” is boring. A “quiet city park in late autumn with orange leaves on the path and a wet bench in the foreground” is believable.
Try to add:
- Time of day, such as early morning, noon, golden hour, or night
- Weather, such as light rain, fog, harsh sun, or snow
- Texture, such as cracked pavement, rusty metal, or foggy glass
- Background life, such as distant traffic, a passing cyclist, or people talking
Background life is a trick that works really well. A scene with ZERO movement behind the subject feels like a still image. A little motion in the back makes it feel alive.
Step 4: Use Camera Language the Model Understands
This is where many realistic prompts win or lose. Veo 3.1 responds well to the same words real filmmakers use. You do not need to be a pro, just learn the basics.
| Camera Term | What It Does | Best Used For |
|---|---|---|
| Close up | Shows face or small object in detail | Emotion, product shots |
| Medium shot | Shows subject from waist up | Conversations, general action |
| Wide shot | Shows full scene and surroundings | Establishing a location |
| Low angle | Camera looks up at subject | Power, scale, drama |
| Over the shoulder | Camera sits behind one person | Dialogue scenes |
| Dolly in / push in | Camera moves slowly toward subject | Building tension |
| Tracking shot | Camera follows moving subject | Walking, running, driving |
| Handheld | Slight natural shake | Raw, documentary feel |
| Static shot | Camera does not move | Calm, stable scenes |
| Shallow depth of field | Background is blurred | Cinematic, subject focus |
One rule: pick ONE camera movement per prompt. If you ask for a pan, a zoom, and a crane shot together, the result often looks confused.
Want a quick example? “Handheld tracking shot following a runner along a rainy street at night, shallow depth of field.” That single line tells Veo the movement, the feel, and the focus.
Step 5: Control Lighting for Believable Results
Lighting is the secret behind most “wow, is this real?” clips. Bad lighting makes AI video look plastic. Good lighting adds depth right away.
Useful lighting phrases include:
- Golden hour sunlight for warm, soft, flattering scenes
- Overcast soft light for calm, natural, low contrast looks
- Harsh midday sun for strong shadows and a gritty feel
- Neon signs reflecting on wet pavement for night city scenes
- Window light from the left for indoor portraits
- Candlelight or warm lamp glow for intimate scenes
Be specific about direction when it matters. “Soft window light from the left, casting gentle shadows on her face” gives a more real result than just “good lighting.”
Step 6: Add Audio, Because Veo 3.1 Can Hear Your Prompt
Native audio is one of the biggest strengths of this model, yet many users skip it. Do not skip it. Sound makes a clip feel finished.
You can describe three types of audio:
- Ambient sound: wind, traffic, crowd noise, rain, or a quiet room
- Sound effects: footsteps, a door closing, glass clinking, an engine starting
- Dialogue: short spoken lines placed in quotation marks
For dialogue, keep it short. Something like this works well:
A young woman looks into the camera and says, “I did not expect it to work this fast.” Quiet cafe ambience in the background, soft clinking of cups.
Long speeches are risky. Lip sync and timing are more reliable with one or two short sentences. Also say who is speaking if there are two people, so the model does not mix them up.
Step 7: Pick a Style, But Do Not Overload It
Style words set the overall look. For realism, you want words that point to real footage, not painted or animated looks.
Good realism style words:
- Natural documentary footage
- Cinematic, 35mm film look
- Smartphone video, vertical, handheld
- News report style
- Security camera footage
- Product commercial, clean studio lighting
Here is a mistake I see a lot. People write “ultra realistic, 8K, masterpiece, hyper detailed, best quality” at the end of every prompt. Does this help? Barely. These are old image prompt habits. Veo 3.1 gives more value to real scene description than to a pile of quality buzzwords. Spend those words on lighting and camera instead.
Full Prompt Examples You Can Copy
Below are three complete prompts using the formula. Change the details and make them your own.
Example 1: Cinematic Street Scene
A woman in her thirties wearing a red raincoat walks slowly down a narrow city street at night, holding a clear umbrella. Tracking shot from behind, then she turns her head slightly toward the camera. Neon shop signs reflect on the wet pavement, light rain falling. Cinematic look, shallow depth of field. Sound of rain on the umbrella, distant traffic, and soft footsteps in puddles.
Example 2: Product Commercial
A glass bottle of cold sparkling water sits on a wooden table in a bright modern kitchen. Slow push in, close up, as a drop of condensation runs down the glass. Clean morning window light, soft shadows. Premium commercial style, sharp focus on the bottle. Sound of a gentle fizz and a quiet kitchen ambience.
Example 3: Nature Documentary
A red fox stands at the edge of a snowy pine forest at sunrise, ears turning toward a sound. Wide shot, then a slow dolly in. Cold blue shadows with warm orange light on the treetops. Natural wildlife documentary footage. Sound of light wind, crunching snow, and a distant bird call.
Using Reference Images and Frames
Veo 3.1 also supports image based workflows in many tools. You can give a reference image to help keep a character, outfit, or product consistent across clips. Some workflows even let you set a first frame and a last frame so the model fills in the motion between them.
A few simple tips here:
- Use a clear, well lit reference image with the subject easy to see
- Keep your text prompt focused on MOTION and camera, since the image already shows the look
- Reuse the same description of your character across prompts for better consistency
Feature availability can change depending on the platform you use, so check what your tool supports before planning a long project.
Common Mistakes That Make AI Video Look Fake
| Mistake | Why It Hurts | Quick Fix |
|---|---|---|
| Prompt is too short | Model guesses everything | Add camera, light, and audio |
| Too many actions | Clip looks rushed or messy | One scene, one main action |
| Several camera moves | Motion feels confused | Choose a single movement |
| Abstract emotion words | Hard to show on screen | Describe visible behavior |
| Buzzword stacking | Wastes prompt space | Use real scene details |
| Ignoring audio | Clip feels empty | Add ambience and effects |
| Long dialogue | Sync and timing problems | Keep lines short |
| Crowded scenes | Faces and hands break | Use one or two subjects |
Hands, fast motion, and big crowds are still tricky for every AI video model. If your clip has these, keep the shot simple and try a few versions.
A Simple Workflow to Improve Your Prompts
Great results rarely come from the first try. Here is a short process that works well.
- Write a basic version with subject, action, and setting
- Add camera and lighting details
- Add audio lines
- Generate and watch the clip carefully
- Change ONE thing at a time, then generate again
Step five is the most important. If you change the camera, the lighting, and the subject together, you will never know which change helped. Small edits teach you faster.
If you struggle to find the right words, you can use an AI assistant to help you draft and polish prompts. For example, Claude is good at turning a rough idea into a clean, detailed scene description that you can paste straight into your video tool. Just tell it the formula above and ask for three variations.
Negative Details and What to Leave Out
Sometimes you want to avoid certain things, like text on screen or random logos. It helps to describe what you DO want clearly, and then add a short note about what to avoid, such as “no text overlays, no subtitles, no watermarks.”
Keep these notes brief. A long list of “do not” items can confuse the result. Positive description almost always beats a long list of bans.
Quick Checklist Before You Hit Generate
Run through this list every time.
- Does the prompt name a clear subject with details?
- Is there only one main action?
- Did you describe the setting and time of day?
- Is there one camera angle and one movement?
- Did you add lighting direction?
- Did you include audio, such as ambience or a short line?
- Is the style based on real footage, not buzzwords?
If you can say yes to most of these, your prompt is already better than most.
Final Thoughts
Writing good Veo 3.1 prompts is a skill, and like any skill it improves with practice. The model is strong, but it rewards people who describe scenes like a filmmaker. Think about what a real camera would see and what a real microphone would hear. Then write exactly that.
Start with the seven part formula, keep each prompt focused, and change one detail at a time. In a short while you will notice your clips look more natural, more cinematic, and much less like generic AI output.
