Veo 3.1 prompt library: real prompts from our gallery

Browse by category

Looking for a workflow rather than a single prompt? The use-case guides turn these examples into step-by-step plans for product ads, vertical shorts, YouTube videos and explainers, with the credit cost of a typical deliverable.

How we built this library

On November 8, 2026 we pulled all 155 videos in the public gallery. Today the generator only shares a video when its creator switches sharing on, and videos made from uploaded images can't be shared. We read every prompt in full and left out anything not safe for work or violent, prompts naming real people, brands, private businesses or personal details, copyrighted characters and game footage, and prompts about real disasters. From what was left we kept the 47 prompts shown here, including a few that went wrong. If a creator unpublishes a video, its prompt drops off these pages within a day.

The gallery records the model and length of each video, but not always the other settings correctly, so we opened every file and measured its frame size, length and whether it has a sound track. Here is what the whole gallery looked like:

Costs on these pages are today’s price for the measured setting. A video shared in July may have cost a different amount when it was made.

Six things the gallery teaches about Veo 3.1 prompts

1. The controls decide the frame, not the prompt

The soda-bottle clip asks for “Vertical 9:16” and “4K” and came out 1920×1080 landscape. The Neptune macro asks for 8K and 120 fps and is 1080p at 24 fps. Set aspect ratio, resolution and duration in the generator, and spend the prompt on what the viewer sees.

2. Sound cues need With Audio switched on

The wet-boots UGC clip ends with careful sound direction (rain, a ticking radiator, a sigh), and the file is silent: either audio was off or the model returned no sound. Audio off is cheaper, so it’s the right choice for b-roll you will put music or voiceover over. Just don’t write sound you aren’t paying for.

3. Describe the action, not only the look

The miniature chef prompt spends almost 1,500 characters on props, lenses and lighting and never says what the chef does. Strong prompts in the gallery state one clear action and one camera move, then add style.

4. Anything you leave open, the model decides

The same cartographer prompt gave an older man on Fast and a woman on Lite. If a character has to match across clips, describe their age, look and clothes the same way every time.

5. On-screen text is close, not exact

The anatomy model asked for the caption “One body. Every system, working together.” and rendered “One body Every system, working together.” That’s usable, but not exact. Add text that must be right in your editor; our page on garbled subtitles explains how to keep it out of the frame.

6. Write for 8 seconds

Several gallery prompts plan 10 or 15 seconds or a 30-second script. A clip is 4, 6 or 8 seconds, so write one scene per clip and join clips in an editor. Our 8-second limit guide covers stitching, and the prompting guide covers prompt structure.

A few to start with

One example from each kind of prompt in the library. The category pages have the full prompts with copy buttons.

• Miniature anatomical model with a caption

  Veo 3.1 Fast · 9:16 · 1080p · 8 s · audio on · 10 credits (about $2.13–$2.50)

  It has one strong idea (a transparent body model), a camera plan (“slow orbiting camera reveal”) and an order for the highlights: heart, lungs, digestive system, brain. In our frame grabs the caption rendered almost word for word (“One body Every system, working together.”). That is close, but the first period is missing, which is typical of generated text.

• Product-problem hook: soaked shoes after work

  Veo 3.1 Fast · 9:16 · 1080p · 8 s · audio off · 6 credits (about $1.28–$1.50)

  This is a classic ad opening in three timed beats: arrival, a reaction close-up, then the problem revealed. Timestamps give the model an order of events. Our frame grabs move from the man’s face to a close-up of wet shoes, as written.

• Night-time creator studio with a music cue

  Veo 3.1 Lite · 9:16 · 1080p · 8 s · audio on · 5 credits (about $1.07–$1.25)

  It splits picture, sound and exclusions into three blocks. The audio block reads like a music brief: a boom on the first frame, a rising synth, a hit at the end. It adds two foley cues (a shutter click and a whoosh), and “No spoken dialogue, no voiceover” stops the model from adding a narrator. Our frame grabs follow the shot list: camera, softbox, laptop and steaming coffee, then the window and the skyline.

• A cartographer’s line, on Fast at 1080p

  Veo 3.1 Fast · 16:9 · 1080p · 8 s · audio on · 10 credits (about $2.13–$2.50)

  This is the simplest dialogue pattern that works: set the scene, then write Speaker: “line”. The line is short enough to fit in 8 seconds. Our frame grabs show an older man leaning over the map in a lamplit study.

• Documentary narration in Turkish

  Veo 3.1 Fast · 16:9 · 1080p · 8 s · audio on · 10 credits (about $2.13–$2.50)

  It names the voice (calm male documentary narrator), the language and pronunciation, the room sound (archive room tone, paper rustling, a ticking clock) and says “no music”. The narration is one sentence.

• Cross-section of a tooth cavity (clinical explainer)

  Veo 3.1 Fast · 9:16 · 1080p · 8 s · audio off · 6 credits (about $1.28–$1.50)

  In about 20 words it gives the subject, the process (“progressing through enamel into dentin”), the style and the background. A white background makes the clip easy to drop into slides or an explainer edit.

Cheapest example in the library: 1 credit (about $0.21–$0.25). Most expensive: 10 credits (about $2.13–$2.50). If you’re unsure which model to use, our Lite vs Fast vs Quality comparison explains the trade-offs.

Frequently asked questions