Google Opens Gemini Omni 1.1 Video Generation to Every Google Account

Google Vids opened Omni 1.1 to every Google account on September 23

Google announced on September 23, 2026 that anyone with a Google or Google Workspace account can create AI video in Google Vids at no cost with Gemini Omni 1.1 Flash. On desktop, users can open vids.new and select “Create AI videos.” This is a material access expansion for a current AI-video product, not a new base-model debut: Google DeepMind originally introduced Omni 1.1 for developers on August 27.

The announcement says the free Vids route can generate new scenes in 1080p HD, upscale existing AI clips, extend scenes while maintaining visual context and set specific clip durations. Google presents the feature as available now. Gemini 3.8 Flash-Lite text-to-speech narration across more than 100 languages is described separately as coming soon, so it should not be counted as part of current availability.

The free product route has no published numeric allowance

Google repeatedly says users can start at no cost, but the reviewed launch post does not state how many generations, seconds, credits or retries a free account receives. It says Google AI plans provide “more access” for personal accounts and that Workspace Business and Enterprise plans include expanded generation pools and centralized administration. That establishes tiering, not a usable capacity forecast.

Teams should therefore verify the allowance inside each intended account before promising campaign volume or turnaround time. Record the plan, country, displayed credits or limits, reset period, queue behavior and what happens after exhaustion. A free entry point is useful for evaluation; an unspecified pool is not an SLA.

The Gemini API is paid and uses a different contract

Developers integrate the generally available model as gemini-omni-1.1-flash through the Gemini API and Interactions API. Google’s pricing page lists no API free tier for this model. Standard paid rates are $1.50 per million text, image, video or audio input tokens, $9 per million text-output tokens and $17.50 per million video-output tokens.

Google says 720p output is billed at 5,792 tokens per second, making the listed video rate approximately $0.10 per second. That is about $1 for ten generated seconds before retries, alternative takes, storage, review or downstream editing. The Vids “at no cost” offer and the API price are not contradictory; they are different access products with different controls and economics.

Omni 1.1 covers generation, references, editing and extension

The current API documentation supports text-to-video and image-to-video generation, first-and-last-frame interpolation, reference images and videos, and stateful follow-up editing using a previous interaction ID. The model accepts multimodal inputs and returns MP4 video, either inline or through URI delivery for outputs larger than 4 MB. The default documented resolution is 720p, with resolution controls exposed in the response format.

For extension, the model generates a 3–10 second continuation and can extend generated scenes in multiple turns up to 40 seconds total. It uses up to the last ten seconds as context. Those controls may reduce manual assembly, but Google’s consistency and narrative-quality descriptions are vendor claims; buyers still need workload-specific review across characters, products, text, audio and cuts.

Several limits matter before production

Uploaded source videos used for editing or extension must be ten seconds or shorter. Extension appends only at the end; it cannot prepend footage or insert into the middle. The API cannot add new dialogue while extending an uploaded clip in which someone is already talking, audio-reference uploads are unsupported, and reference-video audio is ignored. Multiple-video reasoning is also unsupported.

Editing or extending uploaded videos is unavailable in the European Economic Area, Switzerland and the United Kingdom, although Google says extension of model-generated video remains supported in available regions. System instructions, temperature, top_p, stop sequences and dedicated negative-prompt fields are not supported. These are product-contract constraints, not minor prompt tips; test the intended region, source media and edit path before committing a workflow.

SynthID helps provenance but does not settle publication rights

Google says every Omni 1.1 clip generated in Vids and every API-generated video contains an imperceptible SynthID watermark. The API documentation says the watermark can be detected programmatically. AccessAllGPT did not generate a clip, run a detector or test whether common exports, edits, transcodes or platform uploads preserve a detectable signal.

A production workflow still needs source and consent records, rights review for reference media, visible disclosure where context requires it, an approval owner and a response path for impersonation or disputed content. Treat SynthID as one provenance signal, not proof of ownership, permission, factuality or harmlessness.

Start with Vids, then earn the API integration

Use the free Vids route to test a fixed set of real briefs before writing integration code. Include product shots, people, readable text, camera motion, scene extension and difficult transitions. Record prompt, reference assets, output duration, wait time, retries, accepted result and reviewer reason. Count accepted seconds rather than attractive demo clips.

Move to the API only when automation, stateful editing or volume justifies engineering and paid generation. Re-run the same fixtures through gemini-omni-1.1-flash, calculate total generated cost per accepted second, test regional behavior and preserve every input’s authority. Wait when the account allowance is unknown, a required edit is unsupported or the rights chain is incomplete; reject unattended publishing when a person cannot review the final media and provenance.

Disclosures

AccessAllGPT did not receive Google access, credits, a subscription, API account, briefing, test media, review or compensation for this article. Google and Google DeepMind did not sponsor, review or endorse it. AccessAllGPT Research is operated by NeuralArc, is independent, and is not affiliated with Google, Google DeepMind, OpenAI or organizations cited. Publication-wide relationships are listed on the disclosures page.