Google brought its Omni video model to Vids, its video creation app for work, in July. The new version, Omni 1.2 Flash, adds a slider that sets each clip's length anywhere from 3 to 15 seconds; Google's developer announcement says Omni 1.1 topped out at 10 seconds. Google also says generations that use reference images keep characters and products more consistent, that motion and camera moves show less drift or morphing, and that voices are more varied and realistic.

In Vids you give Omni a written prompt and, optionally, reference images or an avatar, and it returns a short clip with sound. Rollout to both release tracks, Rapid Release and Scheduled Release, began November 5, and Google said the feature could take one to three days to appear. Omni has no admin setting of its own, so who can use it depends on the plan and on the Vids and Gemini settings covered below.

Who gets it and how the allowance works

Eligible plans include Business and Enterprise Starter, Standard and Plus, Enterprise Essentials and Enterprise Essentials Plus, Education Plus, Nonprofits and Workspace Individual, as well as personal accounts with Google AI Pro or Ultra. Frontline plans don't include Gemini features in Vids.

Generation is metered in seconds of video per user per month. Google's availability page lists 200 seconds on Business Starter, Enterprise Starter, Nonprofits and Education Plus, and 500 on Business and Enterprise Standard and Plus, Enterprise Essentials and Enterprise Essentials Plus. The AI Expanded Access add-on raises that to 2,000, and Workspace Individual gets 60.

Allowances belong to each user and can't be shared or carried into the next month. Monthly limits reset at 12 AM Pacific time on the first of each month. A prompt counts once it generates a response, and Google says follow-ups, such as a request to try again, count too, so every redo draws on the same allowance.

The feature's help article adds that most users can generate up to 50 videos a month, so plan against whichever limit you would reach first. At 15 seconds a clip, 500 seconds covers 33 generations and 200 covers 13. Test a prompt at three to five seconds, generate the full length once the short version looks right, and review the long version just as carefully.

Admins can buy AI Expanded Access for specific users or organizational units. Because seconds can't be pooled, buy it only for the people who actually make videos.

Decide whether a shot should be generated

The simplest trial is a short supporting shot in a training or explainer video, such as a section opener, a transition, or a visual for an idea that is hard to film. Give the clip one job, and keep instructions, product claims and anything customer-specific in narration, text or footage your team can verify.

Before generating anything, sort each planned shot:

  • Generate it if it sets a mood, bridges two sections or illustrates an idea, and nobody will act on what it shows.
  • Use the Record option in Vids if it shows your software, so viewers see your real screens.
  • Film or upload it if it shows your premises, staff, products or a step people will follow.
  • Film or simplify it if it needs fast sports action, rapid script changes or complex physics, which Google says it is still working to improve.

Make a clip step by step

On a computer, the steps are:

  • Open Vids and select Create AI videos from the start menu, or AI video on the right side of an existing video.
  • In the side panel, select Create and describe the shot, including how to use any ingredients.
  • Optionally, use Ingredients for up to seven reference images, or Avatar for an avatar with its voice.
  • Select the Omni pill to choose Landscape (16:9) or Portrait (9:16), a length from 3 to 15 seconds on the duration slider, and 720p or 1080p.
  • Select Generate, then edit the prompt and generate again if needed, or select Recreate to reuse it.
  • Select Insert below the clip you want; it gets its own track at the bottom of your Vid.

Clips come out at 24 frames per second. Session history shows the current session's prompts and clips so you can compare and reuse them, but it disappears when you close the Vids tab, so insert anything worth keeping first.

The panel has three other modes. Animate turns an uploaded .jpg or .png image into a moving clip; Google says a sharp, clear image works best and the prompt should describe how things or the camera move rather than describe the scene again. Vids adds black bars when the image's shape doesn't match the nearest supported aspect ratio.

Edit reworks an uploaded video with up to three reference images or an avatar, at 3 to 10 seconds, and only the last 10 seconds of a longer upload are edited. Uploading videos to Edit isn't available in the European Economic Area, the United Kingdom, Switzerland, Illinois or Texas, where users can edit only AI-generated videos in their generation history. Extend continues a clip generated in the current session, using a prompt that describes what happens next.

Write prompts that need fewer retries

Google recommends a clear, precise prompt with rich descriptions and cinematic language, covering the subject, location, action, camera, lighting, dialogue, sounds and tone. For example: slow push-in toward a tidy reception desk in a small dental office, soft morning light, no people, quiet room tone. Add titles and labels with Vids text rather than asking the model to draw words.

When a draft misses, change one element at a time, such as the camera move or the lighting, so you learn which words made the difference. Save prompts that work in a shared document so the next video starts from them.

Use reference images and avatars you have the right to use

Consistency matters most with Ingredients, where a recurring product or character should stay recognizable across clips. Before uploading an image, confirm your organization owns it or has permission to use it this way. Get a person's agreement before using their photo, and never upload screenshots of customer records.

Avatars work as ingredients only in English and on listed plans, including Business, Enterprise, Essentials and Nonprofits. A personal avatar is a likeness of one employee, and Google says the account owner must create it: they select Personal avatar from the Vids start menu, scan a QR code with a phone or tablet, and record their face and voice. The person must be 18 or older, personal avatars are supported only in English, and they aren't available in the EEA, Switzerland or the United Kingdom.

Admins can allow or block them with the Allow personal avatars in Google Vids checkbox in the Vids settings. Set your rule before anyone records one: what an avatar may say, who approves its use, and what happens to it when the person leaves. Deletion happens from the person's Google Account and also affects other services that use the avatar, such as the Gemini app.

Google's Generative AI Prohibited Use Policy applies. It bans violating privacy or intellectual property rights, including using personal data or biometrics without legally required consent, impersonating someone without explicit disclosure in order to deceive, and claiming generated content was made solely by a human in order to deceive. The help article also says generated images and videos are for use only within Vids, so treat clips as parts of a Vids project, not stock footage for other tools.

Review every clip before anyone else sees it

Generated footage can look plausible while being wrong, and Google cautions that Gemini features may suggest inaccurate or inappropriate information. Have someone other than the person who generated each clip watch it full screen with sound on and check:

  • Words on signs, labels and screens.
  • Logos, yours or anyone else's.
  • Faces resembling real employees, customers or public figures.
  • Equipment or software you don't use.
  • Product behavior or results you can't back up.
  • Speech or sound effects nobody approved, and lips that don't match the words.
  • Objects or people that drift, morph or jump between moments.
  • Whether a referenced product or character stayed consistent.

If a clip comes out silent, Google says to check your computer speakers and the speaker icon on the clip's track before regenerating. If the picture is right but the sound is not, muting that track is cheaper than spending seconds on a new generation. The editor still decides whether a generated clip fits the message.

Label generated footage yourself

Google marks Omni output in two hidden ways. Its developer announcement says outputs carry an imperceptible SynthID watermark, and Gemini-generated media in supported Workspace apps, including Vids, carries Content Credentials, embedded provenance information based on the C2PA standard. In a computer browser, right-click the clip and open the Content credentials sidebar; an app listed as Google Media Processing Services means AI made or changed the file.

Content Credentials are fragile. Screenshots, resizing or re-saving in other editing tools, downloads, exports and some social platforms can strip or change them, and client-side encrypted or offline files don't get them at all. Google says SynthID is preserved during most edits, downloads and exports, but neither marker is visible, and Content Credentials can't tell you whether a video is real or fake.

Assume viewers can check anyway. Anyone with a personal Google account or a qualifying Workspace account can upload a video shorter than 90 seconds and no larger than 100 MB to the Gemini app and ask whether it was created or edited by Google AI; for video, Gemini says which parts contain the watermark. If viewers need to know a shot was generated, say so on screen or where you share it.

Verify the finished video

To get an MP4, select Quick sharing options next to Share and then Export to Drive. Export to YouTube, under the File menu, uploads to your default channel with visibility set to Private. Before anything goes out, confirm:

  • Each generated clip passed the checklist with a named reviewer.
  • The Content credentials sidebar shows Google Media Processing Services on every clip you think is generated.
  • The disclosure appears on screen or in the sharing message, not only in metadata.
  • The exported MP4 plays end to end with titles and disclosure intact.
  • People who only need to watch have Viewer access, not Editor.

Check access before you troubleshoot

Vids itself has a switch. An admin with the Service Settings privilege goes to Menu, then Apps, then Google Workspace, then Drive and Docs, then Google Vids, and sets it On or Off for the organization, an organizational unit or an access group. Changes can take up to 24 hours. Users with Vids off can't create Vids files or open ones shared with them, including files from other organizations.

On Enterprise Standard and Enterprise Plus, an admin with the Gemini Settings privilege can also turn Gemini off under Menu, then Generative AI, then Gemini for Workspace, then Feature access. It can be set by organizational unit or configuration group, and group settings override organizational units. Google's help lists Vids in one entry with Drive, Docs, Sheets, Slides, Forms and Drawings, so check what else a change covers before you save.

If someone can't find Create AI videos, check in this order:

  • Their license is on an eligible plan, not Frontline.
  • Vids is On for their organizational unit or access group, with 24 hours allowed after any change.
  • On Enterprise plans with Feature access, Gemini is on for them.
  • Smart features are on, which Google says users need for Gemini in Workspace services; in Gmail they are under Settings, See all settings, General.
  • They use a computer with one of the two latest versions of Chrome or Firefox, or Microsoft Edge on Windows.

The AI options appear in the start menu only for accounts with AI access in Vids. Safari can't edit or comment in Vids, and mobile devices can play processed videos but can't edit them.

Common mistakes and how to spot them

Signs that something went wrong, and what they usually mean:

  • A session history full of 15-second attempts means drafts are running at full length.
  • The same prompt three times in a row means someone is retrying instead of rewriting.
  • An empty history after a break means the tab was closed, and anything not inserted is no longer in the session history.
  • Garbled lettering on a sign or screen means text crept into the shot; prompt it out and add words with Vids text.
  • A viewer asking where to find something shown in a generated shot means the shot is doing a job real footage should do.
  • A forwarded MP4 with nothing on screen about AI means the disclosure lived only in metadata or the share message.
  • Staff suddenly unable to open a partner's Vids file can mean Vids was switched off for their organizational unit.
  • Feedback sent through Help Vids improve that names a client is a privacy slip; Google asks for no personal, confidential or sensitive information there.

A worked example

Imagine a twelve-person property management office making a three-minute onboarding video about logging maintenance requests. The office is on Business Standard, so the office manager has 500 seconds this month. The software steps are a narrated screen recording made with the Record option in Vids, and she plans two generated shots: a six-second building exterior to open and a four-second transition before the section on emergencies.

She drafts each shot four times at three seconds. When a transition draft shows a technician holding a tool the company doesn't use, she rewrites it as an empty hallway with no people. The full-length exterior shows a garbled sign, so she drops signage from the prompt, regenerates it and adds the company name as Vids text.

That is 40 of her 500 seconds in 11 generations: eight three-second drafts, two six-second exteriors and one four-second transition. Had every attempt run 15 seconds, the same 11 would have used 165.

A colleague watches both shots full screen with the checklist and notices the transition's generated room tone is louder than the narration, so the office manager mutes that clip's track. She confirms both clips show Google Media Processing Services in the Content credentials sidebar and adds the words "AI-generated shot" as Vids text in a corner of each, so the note survives an export. She shares the Vids file with staff as Viewers and says in the sharing message that two shots are AI-generated. Her unused 460 seconds won't carry into next month.

Cost and privacy questions

Clips don't cost extra on an eligible plan. Google's developer announcement says 15-second clips come at no additional cost in Vids with a Google account or Google Workspace subscription, and an admin who wants to raise a user's allowance can buy the AI Expanded Access add-on for that user.

Workspace content isn't used to train models outside your organization without permission. Google's Workspace privacy hub says that, on qualifying editions, your content is not human reviewed or otherwise used for generative AI model training outside your domain without permission.

Admins can review Gemini activity. The same hub lists Vids among the Workspace apps covered by Gemini audit logs, and an admin with the Audit & Investigation privilege finds Gemini for Workspace log events under Reporting, then Audit and investigation.

Plan on English. Google's availability page says many AI features in Vids are only available in English at this time.

What to do this week

  • Confirm each video maker's plan and monthly seconds on Google's availability page.
  • Check the Vids setting for their organizational unit, and Feature access on Enterprise plans.
  • Pick one internal video and one job for a generated clip in it.
  • List the images staff may use as references, and decide whether personal avatars are allowed.
  • Name who reviews clips and give them the checklist above.
  • Decide when and how your organization discloses generated footage.

Where ALCO can help

ALCO's Google Workspace service runs the admin console for clients, including users, groups, aliases, licenses and billing, the same console that holds the Vids setting. ALCO's user training covers Google Workspace the way your team uses it. The ALCO help desk is staffed 24 hours a day, including holidays, and calls are answered by a person.

Sources