Edit video with AI in Google Flow: upload, trim to 10 seconds, prompt, and what each edit costs

To edit video with AI in Google Flow, set the model to Omni Flash, upload a clip of up to 60 seconds, select a segment of up to 10 seconds, and type the change you want. Each edit costs 40 credits, Flow keeps context for up to three conversational turns, and Google lists uploaded-video editing as off in the UK, EU and Switzerland.

We read Google's Flow Help pages on October 28, 2026, and every limit below comes from them unless we say otherwise. Those pages change often and do not always agree with each other, so the section on conflicts matters as much as the steps. Upfynd lists 219 AI image and video tools as of October 28, 2026, and Google Flow is the one this guide covers.

Check three things before you upload anything

First, location. Google's Flow Help page on supported regions says "Editing of uploaded videos is currently not supported in the following regions: European Economic Area (EEA), Switzerland, United Kingdom." The same page says Flow never supports editing uploaded videos of identifiable minors or generating videos of prominent people, in any region. Google's getting-started page also says VPNs do not enable Flow in unsupported regions, so we would not expect one to switch this feature on.

Second, account. The same page says you must be 18 or older, be in a supported region, and be subscribed to Google AI Plus, Pro or Ultra. Qualifying Workspace plans get Flow at no extra charge with 50 daily credits. Whether a free account can edit at all is one of the places Google's pages disagree, covered below.

Third, browser. Google recommends a Chromium-based browser such as Chrome or Edge on a desktop computer, and says the most powerful editing features are on a computer rather than the mobile app.

Edit video with AI in Flow: six steps from upload to download

These steps follow Google's Edit videos and build scenes page. Google notes the feature is only available in certain countries.

1. Open Google Flow in Chrome or Edge and click an existing project.
2. At the bottom of the prompt box, open your generation settings, click Video, then pick the video model Omni Flash. Choose orientation, number of outputs and generation length if you want.
3. Upload or drag in a video of up to 60 seconds and 1 GB in .mov, .mp4, .avi or .wmv. If it is longer than 30 seconds, Flow makes you trim it down to 30 seconds first.
4. Click the video. In the trimming window, select a segment of up to 10 seconds.
5. Describe the change in the prompt box and click Generate. Google's own examples are "Change the lighting to a cinematic sunset" and "Add a text overlay that says 'Coming Soon'". You can add ingredients (reference images) to the prompt too.
6. Refine with follow-up prompts. Google says you can continue for up to three conversational turns without losing the context of your previous edits. Every version stays in the History panel, and you can save one to your project and arrange clips in Scenebuilder to download a scene.

Google also describes a second route: turn on the Flow Agent in the prompt box, drag in the video, describe the change and click Generate. The help page does not say whether the Agent route changes the cost, so check the credit figure in your prompt box settings before using either one.

One consequence worth planning for: because a single edit takes a segment of up to 10 seconds, a 30-second clip is, as we read the page, three separate edits. At 40 credits each, that is 120 credits for one pass over the whole clip.

Write the prompt as one change, then say what stays

Google's Gemini Omni announcement shows what a good edit prompt looks like. The multi-turn example is a violinist clip edited in three steps: "Transport the violinist to the image environment", then "Make the violin invisible", then "Change the camera angle to be over the violinist's shoulder." Each prompt changes one thing.

Two of Google's own example prompts also tell the model what to leave alone: "edit this keeping everything the same. add animated motion effects coming out of the skateboard" and "...keeping the room structure the same". Zapier's Gemini Omni guide, published October 26, 2026, gives the same advice in plain terms: keep edit prompts short and specific, because overly detailed requests can cause unintended changes, and add "Keep everything else the same" when you only want to change a camera angle or an outfit. That phrase comes from Zapier's guide, not from Google's help pages.

Our reading of these sources, as opinion: write the change first, the preserved parts second, and resist stacking three changes into one prompt. With a three-turn limit and a 40-credit price, in our view a vague prompt is the expensive mistake.

If your edit needs the same character across several clips, build reference images first. We covered that workflow in AI character consistency.

Each edit costs 40 credits, so budget by edits

Google's credits page lists a flat price for editing, and says the figures are per generation, not per request: a single request can produce two videos, so it can cost two generations.

Google's table is as of October 28, 2026, and it warns that limits can change at any time. On the supply side, everyone gets 50 credits a day, unused daily credits do not roll over, and the daily refresh starts with your first generation. Google AI Plus adds 200 credits a month, Pro adds 1,000, and the two Ultra plans add 10,000 and 25,000. Monthly credits also do not roll over.

Our arithmetic from those figures: 50 daily credits cover one edit with 10 left over, 200 monthly Plus credits cover five edits, and 1,000 Pro credits cover 25. That assumes you spend nothing on generating the starting clip. Google's getting-started page says you are not charged credits for failed generations. We would set the number of outputs to one before each edit, since cost is counted per generation.

Where Google's own pages disagree

Three conflicts matter to anyone planning around Flow's editor. Each is stated as what the pages said on October 28, 2026.

Free access. The getting-started page lists a Google AI Plus, Pro or Ultra subscription as a requirement, and Google's Omni announcement says Omni Flash rolled out to Plus, Pro and Ultra subscribers. The credits page says "With or without a subscription, you receive 50 Google Flow credits per day", and the models page says "All users can access Gemini Omni models." Zapier says anyone on a free personal plan can use Omni. We cannot settle this from here. The credits page does say non-subscribers may be unable to generate videos during peak hours, roughly 2 PM to 5 PM UTC.

Extending a clip. The edit page says "You can currently only extend Veo generated videos." The models page puts "Extend videos" for Gemini Omni Flash 1.1 under "Features coming soon". Zapier's guide describes extending an Omni scene up to 40 seconds. If extension matters to you, look for the Extend button on your own clip before you plan around it.

What a 360p draft saves on an edit. The models page says you can edit or generate in 360p at half the credit cost of 720p. The credits table says generation at 360p is cheaper but lists edits at a flat 40 credits "of all lengths and resolutions". Google's own advice is to check the active model, resolution and latest credit costs in the prompt box settings before you generate. Do that, and treat the on-screen figure as the answer.

When another editor is the better choice

In our view, skip Flow's editor if you are in the UK, EU or Switzerland, if you need to edit more than 10 seconds at a time, or if you only need cuts, captions and trims rather than AI changes to what is in the frame. Several listed tools edit video you upload, and these are descriptions from their Upfynd listings, not test results:

• Runway is listed as an AI platform for generating and editing video and images, with a free tier.
• Kapwing is listed as an online video editor with AI captions, subtitles and text-to-video.
• CapCut is listed as an AI video editor with templates, auto captions and script-to-video.

We have not tested any of these against Flow's editor, and we are not ranking them. Check each vendor's own pricing page for current limits, and browse the full AI Image & Video category to compare more.

Watermarks, rights and what to read yourself

Google's getting-started page says every Flow output made with Veo, Omni or Nano Banana carries an invisible SynthID watermark. A visible watermark can be switched on or off with the "Visible watermarking" toggle under your profile picture, but Google says it is applied automatically for users in India, South Korea and Vietnam who are not on the Google AI Ultra 20x plan.

On ownership, the page quotes the terms: "Some of our services allow you to generate original content. Google won't claim ownership over that content." That is Google's statement about its own claim, not a ruling on copyright in your edited footage, on rights in the source clip, or on what a platform or client will accept. We read these documents on October 28, 2026. Terms and help pages change. This is not legal advice: read the current Google terms before you publish or sell edited footage, and take advice where real money rests on it. Jurisdiction matters here, including your own.

Your next step is cheap: open Flow, run one edit on a 5-second clip, and read the credit figure Flow shows before you press Generate. If you build an AI video tool yourself, you can submit it to Upfynd for a free listing.