Veo 3.1 Tutorial: Make a Product Ad From One Photo, Sound and All

To make a product ad with Veo 3.1, you need one photo and one prompt. Here is the whole tutorial in five steps:

1. Open Veo 3.1 in OpenClips or Google Flow and add your product photo as the first frame.
2. Write the prompt: one camera move, one action and the sound you want to hear.
3. Pick Veo 3.1 Fast for the first takes.
4. Set the frame (16:9 or 9:16), the longest length, the top resolution and sound on.
5. Generate, check the take with the volume up, then render the same settings on Veo 3.1 for the final.

That is exactly what Ana does in our video below with an espresso machine. The rest of this post goes deeper: how to write the audio, Fast vs standard, Flow vs OpenClips, five prompts to copy and the mistakes that waste credits.

Watch: our Veo 3.1 tutorial, Fast vs Veo 3.1. The full run in under three minutes: photo, prompt, settings, price on the button, and both renders played out loud.

Veo 3.1 tutorial: a product ad from one photo, step by step

Step 1: Prepare the photo

Use a sharp shot of the product with the label readable and nothing covering it. A plain or real-life background both work. Crop it to your final frame before you upload: 9:16 for Reels, TikTok and Shorts, 16:9 for YouTube and the web. The photo sets the frame, so a square photo in a vertical ad forces Veo to invent the edges.

Step 2: Choose how Veo uses the photo

• As the first frame (image to video). The ad opens on your exact photo and moves from there. Use it when the packshot must be true at second 0. Our tutorial uses this mode.
• As a reference (Ingredients to Video in Flow, Reference mode on OpenClips). Veo builds a new scene and keeps your product’s look. Use it for lifestyle shots: a hand, a bathroom shelf, a gym bag. Give it a few references, for example product, person and place.
• First and last frame (Frames to Video in Flow). You give the opening shot and the end card, Veo makes the move between them. At the time of writing this is a Flow feature; OpenClips lists text to video, image to video and multi-reference for Veo 3.1.

Step 3: Write the prompt with the sound in it

Three parts, one sentence each: the camera (a slow push-in), the action (espresso streams into the cup and the crema rises) and the sound (the machine’s hum and the soft trickle of coffee). That is the brief from our tutorial, and it is enough.

Name the product by what it looks like (“a frosted glass serum bottle with a gold dropper”), not only by brand. Keep one action per shot; a short clip goes fast. For dialogue, music and longer briefs, use the audio formula below.

Step 4: Pick the model and the settings

On OpenClips: open Generate, click Change, pick Veo 3.1 Fast. Then set widescreen or vertical, the length, the resolution and sound on. On OpenClips the top resolution is tied to the longest length, so set the length first. The price shows on the Generate button before you click. Every press, with pictures, is in the Creator Handbook: How to use Veo.

In Google Flow: choose Veo 3.1 Fast and check the credit cost under the settings icon in the prompt box (20 credits a clip for non-Ultra plans, per Google’s help page).

Step 5: Check the take, then make the final

Watch with sound on and check four things:

• The label and shape still match your photo.
• The camera did the one move you asked for.
• Every sound matches something you can see.
• Nothing was added to the frame that you did not write.

Fix one thing per re-roll. When the Fast take is right, switch the model to Veo 3.1, keep everything else, and generate the final. In our video, the Veo 3.1 hero take plays at 1:57.

Try it on your own product photo: open Veo 3.1 on OpenClips →

Veo 3.1 Fast vs Veo 3.1: which one for ads?

Lengths, resolutions, references and starting prices on OpenClips, straight from the catalog:

We ran the same espresso photo and prompt on both, with the same settings and sound on. Both renders were ready in under 2 minutes. Fast got the whole brief: the slow push-in, both streams of espresso and the crema rising in the glass. Veo 3.1 went further: the camera pushed right up to the cup, the streams glowed gold in the sunlight and the crema settled thick on top. One test is not a benchmark, but the pattern matches the price gap: Fast for the idea, Veo 3.1 for the hero shot.

Watch: the two renders side by side, from 1:29. Fast first, then Veo 3.1 at 1:57, so compare how close the camera gets and how the crema looks.

Our rule: test on Fast, finish on Veo 3.1. Two Fast drafts plus the final cost less than two standard takes, and the prompt carries over unchanged because it is the same model family with the same inputs.

The audio prompt formula, from Google’s own guide

Google’s Veo 3.1 prompting guide gives one formula for the picture and separate tags for the sound.

Picture: [Cinematography] + [Subject] + [Action] + [Context] + [Style & Ambiance]

Google DeepMind’s prompt guide also allows the sound in a separate Audio: line. We use a Music: line as our own convention because it is easy to read and to set to “none”; it is not a Google tag. Here is the reusable block:

[Shot and lens], [product and who holds it], [one action], [place], [light and mood].
[Name] says, "[one short line]".
SFX: [one or two sounds the product makes].
Ambient noise: [the room or street].
Music: [mood and instruments], or "none" if you will add your own track.

Three habits that keep the sound clean:

• One short line of dialogue per clip. A long line gets rushed or cut. Google itself says spoken audio “remains an area of active development”.
• Say it once. Our Veo 3.1 sample prompts on OpenClips add “one single time and never repeating herself” to stop the line from looping.
• Describe what you want, not what you don’t. Google’s guide prefers “a desolate landscape with no buildings or roads” over a bare “no buildings”. The same works for sound: “a quiet room” beats a list of noises to avoid.

More on structure: our video prompt formula.

5 Veo 3.1 product-ad prompts to copy

Prompt 1 is the brief from our tutorial, written out in full; the video shows the result, but this is our write-up of the brief, not the exact text typed in it. Prompts 2 to 5 follow Google’s formula and tags and have not been rendered yet, so treat them as starting points and swap the brackets.

Slow push-in toward the glass cup. Espresso streams from the [home espresso machine] into the empty glass cup and the crema rises.
SFX: the machine's low hum, the soft trickle of coffee into glass.
Music: none.

Slow push-in, macro lens, shallow depth of field. A [frosted glass serum bottle with a gold dropper] stands on a wet marble shelf beside a folded white towel. A hand lifts the dropper and one clear drop falls back into the bottle. Morning window light, soft steam, calm spa mood.
SFX: the glass dropper clicks against the rim, one soft drip.
Ambient noise: a quiet bathroom, a distant tap running.
Music: none.

Handheld selfie framing, phone at arm's length. A woman in her late twenties stands in a sunny kitchen holding a [can of sparkling citrus drink] up to the lens, label facing camera. She smiles and says, one single time, "Okay, this is the only one I actually finish." She opens the can and takes a sip. Warm daylight, casual creator energy.
SFX: the can opens with a sharp hiss.
Ambient noise: a quiet kitchen, birds outside the window.
Music: none.

Low-angle tracking shot at ground level. A runner's feet in [white running shoes with a mint sole] land on wet city pavement at dawn, splashes catching the low sun. The camera rises as she slows and stops, the shoe filling the frame. Cool blue street light warming to gold.
SFX: rubber soles slapping wet concrete, one deep exhale.
Ambient noise: an empty city street, a distant bus.
Music: a low, building electronic pulse.

[00:00-00] Top-down shot. Hands tear open a [kraft paper coffee bag with a red label] on a wooden counter and pour dark beans into a grinder. SFX: paper tearing, beans rattling into metal.
[00:04-00] Close-up, slow push-in. Coffee drips into a glass cup, steam rising, the bag in soft focus behind. A man off camera says, once, "Worth getting up for." Ambient noise: a quiet morning cafe. Music: soft acoustic guitar.

If your ad shows an AI person, as in prompt 3, check each ad platform’s rules on labeling AI-generated content before it runs.

What Veo 3.1 is, and what Google Flow is

Veo 3.1 is Google DeepMind’s video model, launched on 15 October 2025 with “richer audio, more narrative control, and enhanced realism”, together with the cheaper Veo 3.1 Fast. From Google’s docs:

• Frame: 16:9 or 9:16, at 24 fps.
• Resolution: the standard model reaches a higher top resolution than Fast.
• Sound: dialogue, sound effects and ambient noise in the same render.
• Ingredients to Video: reference images of a character, object or scene.
• Frames to Video: a start and an end image, with the motion filled in between.
• Extend: each new clip builds on the last second of the previous one, so a scene can run “a minute or more”.

Google Flow is Google’s AI creative studio for video and images, available in over 140 countries. Since May 2026 it also runs Gemini Omni Flash and an agent for brainstorming and batch edits. Veo 3.1 is in Flow in three versions: Fast, Quality (the standard model) and Lite.

For developers: the Veo 3.1 preview in the Gemini API shut down on 22 October 2026. Google points to Gemini Omni Flash or the GA Veo 3.1 models on its Gemini Enterprise Agent Platform. Veo 3.1 in Flow and on OpenClips is not affected.

Google Flow or OpenClips: what is different

Both run the same Google model. The difference is everything around it.

Pick Flow if you work in Google’s tools and want Extend or Frames to Video. Pick OpenClips if you want Veo next to other video models on one bill, with the price on the button and ad tools around it. The full credit math: how credits work.

When Veo 3.1 is not the right pick

• You need a longer take in one pass. Seedance 2.5 on OpenClips makes much longer takes. See how to make longer AI videos.
• You need many cheap versions. In our three-model test, Kling 3.0 was the lowest price of the three for the same clip. Compare more in Seedance vs Veo vs Sora.
• You need the same face across a series. Veo takes a few references per clip. Character Studio trains a custom avatar from three photos of one person, plus their consent, for a face that recurs in every ad.

Watch: Veo 3.1 vs Kling 3.0 vs Seedance 2.5: Best AI Video Model for Ads?. One perfume photo and one prompt on three models: watch Veo’s crown splash with its own sound at 1:32, then compare Kling’s calm push-in and Seedance’s full one-take story.

Our verdict from that test: Veo 3.1 for one striking hero shot, Kling 3.0 for lots of test versions, Seedance 2.5 for a full commercial with your avatar.

Common mistakes and quick fixes

One photo, one prompt, one take with sound. New to OpenClips? Start free with ⚡50 on signup, no card.

Sources, read on 29 October 2026: our YouTube walkthroughs (the Veo 3.1 tutorial and Veo 3.1 vs Kling 3.0 vs Seedance 2.5), Google Flow Help on managing Flow credits, Google Cloud’s Veo 3.1 model page, Google AI for Developers on generating videos with Veo, the Google Cloud Blog’s “ultimate prompting guide for Veo 3.1”, Google DeepMind’s Veo page and prompt guide, Google’s posts introducing Veo 3.1 in Flow and in the Gemini API, Google’s Flow updates from I/O 2026, and the OpenClips pages for Veo 3.1, Veo 3.1 vs Veo 3.1 Fast, Seedance 2.5, Character Studio and pricing.