How to Use Google Flow and Veo: A Step-by-Step AI Video Guide

Last updated: 13 November 2026

What is Google Flow?

Flow is a creative workspace from Google designed for making short films, ads, social videos and visual concepts with AI. Instead of generating one-off clips in a chat window, Flow organises your work into projects with scenes, shots and assets, closer to how a video editor or storyboard artist works.

Under the hood, Flow uses Google DeepMind’s video models. Veo powered Flow from its launch, with Veo 3 and 3.1 adding generated sound, dialogue and stronger realism. At Google I/O in May 2026, Google introduced Gemini Omni, a multimodal model that can create and edit video from any mix of text, images, audio and video, and brought it into Flow and the Gemini app. In August 2026, Gemini Omni 1.1 Flash added scene extension up to 40 seconds, keyframe control, cheaper drafts and higher-resolution output in Flow.

The result is that Flow now feels less like a novelty generator and more like a practical tool for producing short marketing and social video, with the usual caveat that AI video still needs a human eye for quality and accuracy.

Flow, Veo, Gemini Omni and the Gemini app: what is the difference?

If you want a quick clip, the Gemini app is enough. If you want to build a sequence of consistent shots, edit them and keep everything organised, Flow is the better place.

Who can use Flow and what it costs

Flow is available to subscribers of Google AI Plus, Google AI Pro and Google AI Ultra, with higher tiers getting more generation credits, access to more capable models and fewer limits. Some features, such as Omni avatar videos, roll out gradually or only in certain countries. Google also offers video generation free of charge inside YouTube Shorts and the YouTube Create app, with tighter limits.

Plans, credit allowances and prices change often and vary by country, including India. Check Google’s current plan page before you subscribe, and look at how many credits each type of generation uses, because high-resolution renders and longer clips use more.

Step-by-step: making your first video in Flow

1. Sign in and open Flow. Use the Google account linked to your AI plan. Create a new project and give it a clear name, such as “Diwali offer reel” or “Product launch teaser”.
2. Choose your starting point. You can start from a text prompt, from an image (your product photo, a sketch or a generated frame), or from ingredients such as a character and a setting you have saved.
3. Write a detailed prompt. Describe one shot at a time. See the prompt formula below.
4. Generate a low-resolution draft. Drafts are faster and cheaper, so use them to test ideas before committing credits to a final render.
5. Review and refine. Change one thing at a time: camera angle, lighting, pace or wording. Keep notes of what worked.
6. Extend or add shots. Use scene extension to continue a clip, or add new shots that reuse the same character and setting.
7. Use keyframes for control. Set a first and last frame when you need a precise transition or loop.
8. Render the final version. Upscale chosen shots to 1080p or 4K where your plan allows, then export.
9. Finish in an editor. Add captions, logos, music and voiceover in your usual editing tool, and check the result on a phone.

The prompt formula that works

Video models respond best to prompts that read like a short shot description from a director. A reliable structure is:

Shot type and camera + subject + action + setting + lighting and mood + style + sound.

Put together: “Slow dolly-in at eye level with shallow depth of field on a ceramic chai cup with steam rising, on a wooden table on a sunlit balcony in an Indian city. A hand places a biscuit beside the cup. Warm early-morning light, calm and cosy. Realistic commercial food photography look. Soft birdsong and distant traffic, no music.”

Advanced features worth learning

Scene extension

Scene extension continues an existing clip so you can build a longer moment without starting over. With Gemini Omni 1.1 Flash, Flow can analyse up to the last 10 seconds of a clip, not just its final frame, which helps keep motion, lighting and story consistent. Extensions come in 10-second increments up to around 40 seconds in total. Use it for a product reveal that needs a little more time, a walk-through of a space, or a short narrative beat.

Tip: describe what should happen next rather than repeating the original prompt. “The camera continues to orbit and reveals the logo on the side of the box” works better than pasting the first prompt again.

Keyframe control

Keyframes let you set the first and last frames of a shot, and the model fills in the motion between them. This is useful for:

• Smooth transitions between two product angles.
• Perfect loops for social media backgrounds or website hero videos.
• Before and after shots, such as a messy room becoming tidy.
• Matching your brand’s existing photography at the start and end of a clip.

Ingredients and references

Flow lets you reuse characters, objects and settings across shots. You can upload reference images, and with Omni 1.1 you can provide up to three seconds of another video as a reference to help keep a character or style consistent. For brands, this is the key to making several shots that look like they belong in the same ad.

Draft and final resolutions

Drafts at 360p generate faster and cost a fraction of a standard render, so test pacing, framing and composition at low resolution. When a shot is right, render it at 1080p or 4K. This simple habit can stretch your monthly credits much further.

Conversational editing

Gemini Omni lets you edit video by describing changes in plain language, such as “make it dusk instead of morning” or “remove the second person in the background”. Results vary, so keep the original and compare versions.

Prompt examples for common business uses

For more on AI tools for short-form content, see our guide to the best AI tools for Instagram Reels.

A simple workflow for a 30-second social ad

Here is a practical process for turning an idea into a finished short ad with Flow. It assumes a small business with no film crew, just a product, a brand and a phone.

1. Write a one-line message. What should the viewer remember? For example, “Our cold brew is smooth, strong and ready in the fridge.”
2. Sketch a five-shot storyboard. Opening hook, product close-up, product in use, benefit moment, end card. Write each shot as one sentence.
3. Prepare references. Photograph your real product on a plain background in good light. Upload it as a reference so Flow keeps the packaging accurate.
4. Generate drafts for each shot. Make three or four low-resolution variations per shot and pick the best.
5. Use keyframes for the hero shot. Set the first frame as the closed bottle and the last as the pour into a glass for a precise reveal.
6. Render finals. Upscale only the selected shots.
7. Assemble and brand. In your editor, add your logo, price, offer text, captions and a short voiceover or licensed music track.
8. Test two versions. Run the ad with two different opening hooks and keep the one that holds attention better.

Expect the whole process to take an afternoon once you are familiar with the tools, compared with days for a traditional shoot. The trade-off is less control over fine details, which is why many brands mix AI shots with real product footage.

Google Flow compared with other AI video tools

For a broader comparison now that Sora is gone, read our guide to the best Sora alternatives.

Getting AI video found: YouTube, search and AI answers

Making a good video is only half the job. If you want people to find it through Google, YouTube or AI assistants, a few basics help:

• Publish on YouTube with a clear title and description that say what the video shows and who it is for. YouTube is one of the sources AI assistants frequently draw on; see YouTube for AEO.
• Add captions and a transcript. Text helps search engines and AI systems understand video content.
• Embed key videos on relevant pages of your website, with surrounding text that explains the topic.
• Use VideoObject schema on pages that host important videos. See schema markup for AEO.
• Keep claims accurate. AI assistants increasingly cross-check information, and misleading video can damage trust.

Planning your credits

Credit systems differ by plan and model, and Google updates them regularly, so treat any fixed numbers you see online with caution. A few principles apply regardless:

• Low-resolution drafts cost much less than final renders, so iterate there.
• Longer clips, extensions and 4K output use more credits.
• More advanced models usually cost more per generation than “Fast” or “Flash” variants.
• Set a monthly budget per project and track how many generations each finished second of video actually needs. After a few projects you will know your real cost per video.

If you use several AI subscriptions, our guide to API credits and AI subscriptions explains how to compare value across tools.

Common mistakes and how to fix them

• Too many ideas in one prompt. One shot, one action. Split complex scenes into several shots.
• Vague style words. “Cinematic” alone means little. Describe camera movement, lens, lighting and colour.
• Expecting perfect text. AI video still struggles with readable text and logos. Add text and logos in your editor afterwards.
• Hands and fine detail. Close-ups of hands, fingers and small mechanical parts can look odd. Reframe or change the action.
• Inconsistent characters. Use references and saved ingredients rather than describing the person differently each time.
• Burning credits on final renders. Draft first, render last.
• Skipping the phone check. Most viewers watch on mobile. Check framing, captions and pace on a phone.

Using Flow on mobile

Google announced mobile apps for Flow at I/O 2026, which makes it easier to capture a reference photo on your phone and turn it into a shot on the spot. Mobile is handy for quick social content, idea capture during a shoot, or reviewing drafts while away from your desk. For detailed editing, keyframing and organising many shots, the web version on a larger screen is still more comfortable. Your projects sync across devices when you use the same Google account.

Prompt tweaks that make a big difference

Responsible use: consent, disclosure and brand safety

AI video is powerful enough to mislead, so responsible use matters for both ethics and reputation.

• Get consent for real people. Do not generate recognisable real people, including celebrities, without their permission. Omni’s avatar feature requires you to verify your own likeness for this reason.
• Label AI content. Platforms such as YouTube, Instagram and Meta increasingly require or encourage AI labels on realistic synthetic content. Follow their rules.
• Avoid misleading claims. Do not use AI video to show product results your product cannot deliver.
• Check trademarks and music. Avoid generating other brands’ logos or copyrighted characters.
• Keep records. Save prompts and source files, so you can show how content was made if questions arise.

If your company is writing rules for AI use, our AI usage policy template is a useful starting point.

Who should use Flow, and who should not

Flow is a strong fit for small marketing teams, solo creators, agencies producing concept videos for client pitches, educators building visual explainers, and businesses that need a steady stream of short social content without regular shoots. It is also useful for storyboarding: even if you plan a real shoot, generating rough versions of each shot helps everyone agree on the idea first.

It is a weaker fit when accuracy matters at the level of every detail, such as showing exactly how a product works, regulated industries where claims need careful control, testimonials (which must come from real customers), or content featuring real people who have not consented. In those cases, real footage, edited quickly with tools like Descript, is usually the better choice.

Our take

Flow has matured quickly. With Veo’s realism, Gemini Omni’s conversational editing, scene extension, keyframes and cheaper drafts, it is now a practical everyday tool rather than a demo. The best results still come from people who think like directors: clear ideas, specific shots, patient iteration and honest labelling. Treat Flow as a fast, flexible production assistant, and keep a human firmly in charge of the story and the facts.

Frequently asked questions