# Google Vids with Gemini Omni: how to make free 1080p AI video
Direct answer
Google Vids now lets anyone with a Google or Google Workspace account generate Gemini Omni 1.1 video scenes at no monetary cost on desktop. The free experience supports controllable clip duration, scene extension, transitions and 1080p generation or upscaling. It is not advertised as unlimited: paid Google AI and Workspace plans provide more generation capacity and administration features.
What changed, and what “free” actually means
On 23 September 2026, Google announced that Gemini Omni 1.1 Flash video generation in Google Vids was opening to anyone with a Google Account or Google Workspace account at no cost. The announcement matters because Vids had previously been understood mainly as a Workspace video editor. The new entry point turns it into a broadly available, browser-based AI video tool: visit `vids.new` on a desktop computer, choose Create AI videos, and build generated scenes inside an editable project.
The verified claim is specific. Google says eligible account holders can generate video with Omni 1.1 without paying. Google also tells personal users who need “more access” to consider Google AI plans, while business and enterprise plans receive expanded generation pools and administration. Therefore, the useful interpretation is a real free access tier with capacity boundaries, not unlimited free rendering. Google’s announcement does not publish a universal number of daily clips, and availability can still depend on account, region, age, administrator settings or product rollout.
That distinction protects you from a common search-result trap. Pages targeting “free AI video generator” often mix four different offers: permanent free access, a one-time trial, credits that expire, and a paid product with a free editor but paid generation. Google Vids now qualifies for the first category at the feature level, but Google has not promised unlimited volume. Check the creation screen before planning a campaign around a fixed quantity.
For a broader inventory of genuinely free services, compare the free AI tools directory and the free AI tiers guide. Those pages help separate account-based allowances from promotional credits.
What you can create with Omni 1.1 in Vids
Google’s launch identifies three controls that make this more than a one-shot text-to-video box.
First, Vids can generate new scenes in full 1080p HD and upscale existing AI clips so they fit a consistent timeline. Resolution alone does not guarantee professional output, but it removes the need to pay another service merely to obtain a standard HD export.
Second, you can specify clip duration. Duration control is valuable because a scene has to fit narration, captions and a wider edit. A beautiful eight-second shot is not useful when the voiceover needs twelve seconds, and randomly trimming generated movement often creates an abrupt ending.
Third, the tool can extend a scene with transitions while attempting to preserve visual context, lighting, characters and environment. Continuity is one of the hardest parts of AI video. Extending a successful shot is usually more reliable than asking for a completely new shot and hoping every visual element matches.
Vids also supplies templates and a timeline-based editing environment. Google highlights product launches, local-business promotions, community events, landing pages, social announcements and event signage as example uses. These are good fits because they can be built from short, controlled scenes plus real text, product photography and brand assets. Long dramatic narratives with recurring people, exact physical interactions or legally sensitive factual claims remain harder.
Google says typed text-to-speech with Gemini 3.8 Flash-Lite across more than 100 languages is coming soon. “Coming soon” is not the same as generally available. This guide therefore treats narration as an optional future or account-dependent feature, not a guaranteed part of today’s free workflow.
A zero-cost workflow that avoids wasting generations
The best way to conserve a free allowance is to make decisions before pressing Generate. A useful workflow has six stages.
1. Write one measurable brief
Define the audience, destination, desired action, duration and evidence. For example: “Create a 25-second 16:9 launch video for a local bakery’s website. Show three products from supplied photographs. End with the verified opening date and address. Do not invent prices or customer quotes.”
This brief is stronger than “make a cool bakery video” because it separates facts from creative direction. Put claims such as dates, prices, specifications and locations into normal Vids text layers, where you can inspect them, rather than asking the model to render words into the scene.
2. Convert the brief into a shot list
Plan four to six scenes. Give each scene one purpose, an approximate duration and a transition. A simple product video might use:
- Establishing exterior, three seconds.
- Product close-up, five seconds.
- Preparation detail, five seconds.
- Customer-free interior atmosphere, four seconds.
- Real product grid with factual labels, five seconds.
- Call-to-action card, three seconds.
Short shot lists reduce continuity problems and make regeneration surgical. If scene three fails, you can repair it without rebuilding the whole project.
3. Prompt for observable camera and scene properties
Describe what a viewer should see rather than abstract praise. A robust scene prompt includes subject, action, setting, camera, lighting, palette, motion, duration and exclusions:
> Close-up of a baker placing three sourdough loaves on a wooden cooling rack in a bright neighborhood bakery; slow lateral camera move; soft morning window light; warm natural colors; realistic food photography; five seconds; no logos, no text, no extra hands, no customers.
The exclusions are not magic, but they make the acceptance test explicit. Avoid using living artists’ names, celebrities, trademarked characters or unlicensed brand styles. Use descriptive visual language you can defend and verify.
4. Generate the hardest continuity scene first
Do not begin with the title card. Generate the scene containing the recurring product, person, location or lighting scheme. Once it works, extend it or use it as the visual reference for adjacent shots. This is where Vids’ scene-extension feature can save free capacity: continuity grows from an accepted scene instead of being re-invented from text.
Keep a prompt log in a free document. Record the prompt, duration, result, defect and change. That prevents the common cycle of making several simultaneous changes and losing the version that worked.
5. Edit facts and branding outside the generated pixels
Use the Vids timeline for headings, captions, logos, prices and calls to action. Generated text inside video frames is harder to correct and may contain spelling errors. Supplied product photos are also preferable when identity matters: AI footage can support atmosphere and transitions while real imagery establishes what is actually being sold.
Check color contrast, caption timing, mobile readability and audio levels. Watch once without sound to test visual comprehension, then listen without looking to test narration. Finally, watch the export on the device and page where it will be published.
6. Export, archive and document
Save the final prompt log, source URLs, licenses for uploaded assets, consent records and the exported file together. Record the date because free-tier conditions and model behavior can change. A small evidence folder is much cheaper than reconstructing permission or provenance after publication.
If your project later needs programmatic rendering rather than a browser editor, start with the free AI API guide and verify current quotas independently. A free consumer interface does not imply a free API.
A prompt template you can reuse
Use this compact template for each scene:
Purpose: What this shot communicates. Subject: The visible person, object or place. Action: One clear movement. Setting: Physical environment and time. Camera: Framing, angle and movement. Light and color: Concrete visual treatment. Duration: Exact requested length. Continuity: Elements that must match the prior shot. Exclude: Text, logos, artifacts or unwanted subjects.
For example: “Purpose: introduce a repair service. Subject: a scratched blue bicycle on a clean workshop stand. Action: technician checks the front brake cable. Setting: small daylight workshop. Camera: waist-height medium close-up, slow push-in. Light and color: neutral daylight, blue and graphite palette. Duration: six seconds. Continuity: same bicycle and gloves as previous scene. Exclude: text, visible brand marks, distorted tools, extra fingers.”
Make one change per retry. If composition is right but motion is wrong, change only the action or camera instruction. This gives you useful information about the model and preserves scarce free generations.
Limits to check before committing a project
The public launch page confirms no-cost access and indicates that paid plans provide more generation access, but it does not state one universal free quota. The creation interface is therefore the authoritative place to inspect your current allowance. Do not publish a number copied from an old review as though it were permanent.
Also verify these boundaries:
- Account eligibility: personal Google Accounts are included, but supervised, school or managed accounts may have restrictions.
- Desktop requirement: Google’s launch specifically directs users to `vids.new` on desktop.
- Regional rollout: features may not appear simultaneously in every country or language.
- Administrator controls: a Workspace administrator can restrict services or sharing.
- Storage and export: generated media and projects use Google services; check available storage and the export format needed by your destination.
- Commercial use: product access does not give you rights to someone else’s logo, face, music, footage or confidential information.
- API access: Vids is an application. Its free UI allowance should not be represented as API credits.
If you need predictable high-volume output, price the entire workflow instead of relying on an unspecified free pool. If you only need a handful of launch, community or explainer scenes, the free tier is a practical place to start without entering payment details.
SynthID is useful, but it is not your whole disclosure policy
Google states that every clip generated with Omni 1.1 in Vids contains an imperceptible SynthID digital watermark in the video frames. SynthID is designed to help identify AI-generated media without placing a visible logo over the picture.
That technical marker does not replace visible disclosure when viewers could reasonably mistake the scene for documentary evidence. A synthetic product visualization, reconstructed event, public-service announcement or realistic person may need a plain-language label such as “AI-generated illustrative footage.” Platform rules, advertising law and client policy can require more than an invisible signal.
Watermarking also does not verify that a claim is true. It indicates generation provenance, not factual accuracy, permission or endorsement. Keep four reviews separate:
- Provenance: was the scene generated or edited with AI?
- Rights: do you have permission for uploaded assets, people, music and marks?
- Accuracy: are names, prices, dates and demonstrations true?
- Safety: could the scene mislead, stereotype, impersonate or expose private information?
Google publishes a generative-AI prohibited-use policy, and normal Google terms still apply. Read them for the account and region in use, especially before political, medical, financial, employment or identity-related content.
Quality-control checklist before publishing
Run the exported file through a repeatable review:
- Watch for changing faces, hands, product geometry, shadows, reflections and background text.
- Confirm that cuts preserve direction of movement and do not jump unexpectedly.
- Verify every factual text layer against a primary source.
- Replace generated depictions with real evidence when demonstrating a product result.
- Check that music, fonts, photos and logos have usable licenses.
- Add captions and confirm they remain readable on a phone.
- Provide visible AI disclosure where context requires it.
- Keep SynthID intact; do not present a generated scene as camera footage.
- Ask an uninvolved reviewer what they believe happened in the video. If their interpretation is wrong, revise it.
Accessibility improves performance as well as inclusion. Captions help people watching silently, strong contrast helps small screens, and a logical visual sequence reduces dependence on narration. Include essential information as selectable page text near an embedded video so search engines and assistive technology are not forced to infer it from frames.
Where Google Vids fits among free AI tools
Vids is strongest as an integrated editor for short, scene-based communication. It combines generation, templates and a timeline under one Google account. It is not automatically the best choice for cinematic character consistency, local offline rendering, open-weight experimentation or an automated API pipeline.
Choose it when you want a browser workflow, 1080p scenes, editable business or community content and no initial monetary cost. Choose a local tool when privacy, offline operation or repeatable model control is the priority; the local AI tools directory is a useful starting point. Choose a hosted API only after confirming a genuinely free allowance and cost guard, because video APIs can consume credits quickly.
The most economical workflow may combine tools: write and fact-check the brief in free documents, generate only the visual shots that benefit from Omni, use real photography for identity and proof, edit in Vids, and publish with transcripts and source links. “AI-first” is not the objective. A clear video produced with the fewest uncertain generations is.
Sources and verification date
This guide was checked on 4 October 2026 against Google’s official launch announcement, the Google Vids product page, Google DeepMind’s SynthID overview, and Google’s generative AI prohibited-use policy. Free-tier capacity and rollout details can change; verify the signed-in creation screen before promising a delivery volume or date.
Frequently asked questions
Is Google Vids AI video generation really free?
Google says anyone with a Google or Google Workspace account can generate videos with Omni 1.1 at no cost. It also says paid plans offer more access, so the supported conclusion is free access with limits, not unlimited generation.
Do I need a paid Workspace subscription?
No for the announced personal access: a standard Google Account is eligible. A managed Workspace account can still be affected by administrator settings, age rules, organizational policy and regional rollout.
Can Vids generate 1080p output?
Google says users can generate new AI scenes in full 1080p and upscale existing AI clips. Inspect the exported project as well as the preview, because source assets and downstream platforms can still affect visible quality.
Does SynthID replace an AI-content label?
No. SynthID provides an imperceptible provenance signal. A visible disclosure may still be appropriate or required when realistic generated footage could affect a viewer’s understanding.
Can I call Gemini Omni through a free API because Vids is free?
Do not assume so. Google Vids is a consumer and Workspace application; its no-cost generation access is not evidence of free programmatic API capacity. Verify API pricing and quotas separately.