The best AI video prompts for beginners are short, concrete, and specific, a subject, an action, a setting, and a camera move, written plainly enough that the model has no room to wander. Below are 40 of them, grouped by what you are trying to make, from b-roll to talking-head to product shots. Steal them verbatim, swap in your own subject, and you will have working footage in minutes, and because every generation in CoreReflex saves its model, parameters, and score, any result you like can be reproduced exactly later.
What makes a good AI video prompt
A model cannot read your mind; it can only render what you describe. Strong prompts almost always include five things: the subject (what is in frame), the action (what happens), the setting (where and when), the look (lighting, lens, mood), and the camera move (static, push-in, orbit). Leave one out and the model fills the gap with its average guess, which is exactly how you end up with a generic clip that ignored half your idea.
Keep prompts to one or two sentences. Front-load the most important element. And resist piling on ten adjectives, a few precise ones beat a wall of vague ones. With that in mind, here are the prompts.
B-roll and establishing prompts
- Aerial wide of a city skyline at sunrise, golden light, light haze, slow drone push-in.
- Extreme close-up of espresso pouring into a white cup, steam rising, soft morning light, shallow depth of field.
- Handheld walk through a misty pine forest, dappled sunlight, leaves drifting, gentle bob.
- Close-up of raindrops sliding down a window, blurred city lights behind, moody blue tones.
- Wide modern open-plan office, people working softly out of focus, warm daylight, locked-off.
- Night highway long-exposure, red and white light trails, static wide shot.
- Slow-motion ocean wave curling and breaking, backlit spray, late-afternoon sun.
- Top-down flat-lay of a notebook, pen, and laptop on a wood desk, soft shadows, slow pull-back.
These are your connective tissue, the shots you cut to between talking points. A reliable b-roll library is the backbone of most faceless YouTube Shorts.
Talking-head and UGC prompts
- Medium close-up of a founder speaking directly to camera, soft key light, blurred studio background.
- Handheld selfie-style clip of a creator talking while walking a city street, natural light.
- Two people mid-conversation at a table with microphones, warm lamp light, shallow focus.
- Presenter gesturing at a whiteboard, mid shot, clean even lighting.
- Casual phone-style clip of someone talking while cooking in a bright kitchen, handheld.
- Emotional close-up of a customer speaking, soft window light, neutral background.
- First-person hands opening a product box on a desk, top-down, bright even light.
- Person speaking to a dashboard-mounted camera, daytime, soft window reflection.
Talking-head and UGC styles carry the message; pair them with a script. Our testimonial video script templates plug straight into prompts 14 and 9.
Product and ecommerce prompts
- Sleek smartphone rotating 360 degrees on a smooth white backdrop, studio softbox light, slow spin.
- Low-angle close-up of a running shoe on concrete, dramatic side light, slow dolly-in.
- Slow-motion water droplet hitting a glass serum bottle, clean white set, macro.
- Extreme close-up of a luxury watch face, light sweeping across the dial, slow pan.
- Overhead of a gourmet burger, steam rising, warm restaurant light, slow push-in.
- Close-up of lipstick swatched on skin, bright beauty light, shallow focus.
- Hands typing on a laptop in a café, screen glow on the face, shallow depth of field, static.
- Model in a flowing coat turning in slow motion, studio backdrop, rim light.
Product shots reward precision about material and light. For dozens more angles, see our roundup of ecommerce product video ideas.
Social and short-form prompts
- Person pointing at on-screen text, bright vertical framing, quick punch-in zoom for emphasis.
- Split-screen transition from a messy desk to an organized one, snappy match cut.
- Fast handheld montage of morning-routine moments, natural light, quick cuts.
- Animated quote text over a soft gradient background, slow drifting motion, minimal.
- Close-up of a surprised expression with a punch-in zoom, bright key light.
- Full-body shot of a person dancing in a colorful room, vertical framing, steady.
- Top-down of hands demonstrating a craft step by step, even light, locked-off.
- Product hidden under a cloth, a hand pulls it away, dramatic reveal lighting.
Vertical, punchy, and built for the scroll. If you coach or teach, adapt these for Instagram Reels.
Cinematic and atmosphere prompts
- Cinematic wide of a rain-soaked neon alley at night, anamorphic flares, slow tracking shot.
- Silhouette walking through a wheat field at golden hour, backlit, slow dolly follow.
- Epic wide of a hiker reaching a summit at dawn, sweeping drone orbit.
- Face lit by a single candle, warm flicker, shallow depth of field, subtle push-in.
- Diver silhouetted against shafts of sunlight underwater, slow drift, blue tones.
- Slow-motion snowfall in a quiet forest, cool light, gentle parallax.
- Figure looking out a high-rise window at dusk, reflection in the glass, slow pull-back.
- 1970s-style film-grain shot of a diner interior, warm tungsten light, static wide.
These lean on mood and motion. The camera moves, tracking, orbit, dolly, are where AI video earns its cinematic feel; for a deeper dive, see our Veo prompt examples and Kling camera move ideas.
Reuse any prompt: provenance you can replay
The catch with prompting is that a great result can feel like a fluke, you got the perfect shot once and can never get it back. CoreReflex closes that gap. Every generation carries a portable trace: the model that produced it, the exact prompt and parameters, and the score it earned at the quality gate. So when prompt 33 lands the perfect neon alley, you can replay it, branch from it, or drop the same recipe into another project and get a matching shot. The mechanics of that trace are spelled out in the documentation.
That turns a prompt from a one-time gamble into a reusable asset. Build a library of the prompts that work for your brand, and each one becomes a repeatable production step rather than a lucky roll. It is also what makes these prompts safe to share across a team, the trace travels with the shot, so anyone can reproduce it.
Frequently asked questions
How do you write a good AI video prompt?
Include five elements: subject, action, setting, look (lighting and lens), and camera move. Keep it to one or two sentences, front-load the most important detail, and use a few precise adjectives rather than a long vague list. Specificity is what stops the model from filling gaps with its generic average.
What is a video generation prompt?
It is the text instruction you give a video model describing the shot you want it to produce. In CoreReflex, each shot in the plan carries its own prompt, so a finished film is really a sequence of prompts, and each one is generated, scored, and assembled into the cut.
Can you reuse an AI video prompt later?
Yes. Because every generation saves its model, parameters, and score as a portable trace, any result you like can be reproduced or branched later, no guesswork about which settings produced it, that is what makes a winning prompt a reusable asset instead of a one-off.
Start with one prompt
You do not need a film-school vocabulary to get good AI video, you need a few concrete prompts and a way to keep the ones that work. These 40 are a starting library; find more in our ideas collection, and let the provenance trace turn your favorites into repeatable shots.
Pick a prompt, describe the shot, and watch it generate, start free, no credit card.