Google’s Gemini Omni Video Tool Lands in India, Free on Shorts

Google has switched on Gemini Omni in India, handing users its newest artificial intelligence (AI) model for turning a typed sentence, a photo or an audio clip into a finished video. The tool reached Indian users of the Gemini app on May 29, less than two weeks after Google unveiled it at its I/O developer conference. It generates and edits clips through a chat-style conversation, so a person describes the change they want and the system rebuilds the scene around that instruction.

Most coverage frames this as Google making video creation simple for the masses. The timing matters more. Five weeks before the India rollout, OpenAI discontinued Sora, the standalone app that was meant to own consumer AI video, and the economics that killed it explain why Google’s free-on-YouTube approach sits on very different ground.

What Google Just Switched On for Indian Users

The model in the India rollout is Gemini Omni Flash, the first release in a wider Omni family that Google introduced on May 19. According to the official Gemini Omni launch announcement, the model is built to create video from almost any combination of inputs and to keep editing it through ordinary language rather than timeline software.

Instead of dragging clips across a desktop editor, a user types a request and the system responds. Ask it to swap a background, change a character’s clothing or extend an action, and it remembers the earlier instructions so the scene stays consistent across multiple turns. Google says the model holds onto continuity, keeping characters and physics steady as edits stack up.

The tool accepts a wider mix of starting material than a typical text-to-video prompt box:

• Plain text describing the scene you want
• Still photos used as a visual reference
• Audio clips to anchor sound or pacing
• Existing video footage to edit or extend

Beyond the standalone Gemini app, Gemini Omni Flash also runs inside Google Flow, the company’s filmmaking workspace, and across YouTube Shorts and the YouTube Create app. That spread is the part of the story that outpaces the feature list.

Sora’s Shutdown Set the Trap Google Walked Into

To see why distribution decides this race, look at what just happened to the company that got there first. OpenAI pulled its consumer Sora app and website offline on April 26, with the developer interface scheduled to follow on September 24, per OpenAI’s Sora discontinuation notice. Sora launched as the product that would define AI filmmaking. It closed roughly eighteen months later.

The reason was money, not quality. Reporting on the shutdown put the numbers in stark relief, and the gap between what the product cost and what it earned was never going to close on its own.

• $15 million a day in compute at peak usage, by reported estimates, with each ten-second clip costing roughly $130 to produce
• $2.1 million in total lifetime revenue across the product’s run
• 66% decline in monthly downloads from the app’s November 2025 high

OpenAI spokesperson Kayla Wood said the company would wind down the consumer app and the application programming interface (API, the developer access layer) to focus on “world simulation research” and robotics. Translated, a brilliant model with no cheap way to reach an audience and no path to pay for itself is a cost center, not a business.

That is the trap. Consumer AI video burns cash with every render, and a freestanding app has to pay for marketing, retention and infrastructure all at once. Google launched Gemini Omni straight into that same expensive market, but it did not arrive empty-handed.

Distribution Is the Moat, and YouTube Is the Castle

Google’s advantage is that it already owns the audience Sora had to buy. The Gemini app crossed 900 million monthly active users in May, up from roughly 400 million a year earlier, and reaches more than 230 countries in over 70 languages. YouTube adds a billion-plus Shorts viewers on top of that. The model did not need to find users; the users were already there.

That is why the pricing choice is the headline most outlets buried. Gemini Omni Flash is available to paying Gemini subscribers on the Plus, Pro and Ultra tiers, but it is also rolling out free of charge to creators inside YouTube Shorts and YouTube Create. Google can subsidize the render cost because a finished Short feeds the same ad machine that funds the rest of the company.

Gemini Omni can create anything from any input, starting with video, and every instruction builds on the last so the scene remembers what came before.

That framing, from Google’s launch post, reads like a product pitch. Underneath it is a distribution thesis. The same week, the company also leaned on cost as a competitive lever through aggressive Gemini price cuts at the same I/O event, and it has been threading AI into YouTube through moves like the new set of AI podcast features rolling out to Premium members first. Put together, the pattern is a company using products it already owns to make the cost of one more AI feature close to a rounding error.

How Gemini Omni Flash Handles a Prompt

The model’s pitch rests on three behaviors that separate it from a basic prompt-to-clip generator. Each is worth understanding before deciding what the tool can and cannot do for you.

What It Builds From a Sentence

Google says Gemini Omni carries an improved intuitive grasp of physical forces, including gravity, kinetic energy and fluid dynamics, so motion in a generated scene behaves more like the real world. It also draws on the broader Gemini model’s knowledge of history, science and culture, which is meant to help it reason about what should happen next in a shot rather than just stitch frames that look plausible.

The conversational memory is the practical hook. Because each edit builds on the previous one, a user can refine a clip over several requests without the characters or setting drifting between versions. Technical detail and demo examples sit in the Gemini Omni model documentation from Google DeepMind.

Where the Limits Sit

The free experience ships with guardrails. Clips made through the Flash tier are capped at 10 seconds at launch, a deployment choice Google describes as a limit on rollout rather than the model’s ceiling. Audio and speech editing are restricted while the company keeps testing them, and the riskiest capability, a personal avatar mode that puts your own voice and likeness into a clip, has been held back as Google works through the safety questions.

Those constraints are not arbitrary. A ten-second cap keeps render costs predictable, and holding back avatars limits the most obvious deepfake vector while the labeling system matures.

How SynthID Marks Every Clip

Every video the model produces carries SynthID, an imperceptible digital watermark embedded in the visual signal that survives ordinary sharing and editing. Viewers cannot see it, but Google Search and the Chrome browser can read it and flag the clip as AI-generated. On YouTube, Omni-made Shorts also pick up an on-screen “AI-generated content” label automatically.

The watermark is Google’s answer to the obvious objection that flooding the internet with cheap synthetic video erodes trust. Whether the label survives contact with screen-recording and re-uploads is the open question, and it is one regulators across several markets are now asking.

Gemini Omni Against Veo, Runway and Kling

Gemini Omni does not enter an empty field. Google’s own Veo line already leads several quality rankings, and a cluster of well-funded rivals split the rest of the market on price and control. The split below shows why Google is fighting on access rather than raw output scores.

The takeaway is that Google can afford to give a capable model away because it already runs the platform where the output gets watched. A standalone rival charging seven cents a second has no equivalent place to recoup that cost, which is exactly the bind that closed Sora.

The Bill Google Hasn’t Paid Yet

None of this means the win is clean. The compute math that broke Sora does not vanish because Google has deeper pockets; it just gets absorbed into a larger ledger, and a free tool aimed at hundreds of millions of users multiplies the number of renders Google has to pay for. Subsidy is a choice the company can reverse the moment the strategy stops paying off.

There is also the content problem. YouTube has already warned that mass-produced, copy-pasted AI clips with no added value will be treated as “inauthentic” and can lose monetization, so the same tool that lowers the barrier to making a Short also lowers it for spam. Creators chasing the Shorts fund will test that line hard.

Then there is provenance. SynthID watermarks help, but the broader risk of convincing synthetic media is already visible in cases like an AI voice-cloning scam that cost one family thousands of dollars. Cheaper, faster video generation widens the surface for that kind of abuse even when the legitimate use case is harmless fun.

If the free Shorts integration drives a wave of watchable, well-labeled clips, Google will have proven that distribution, not model size, decides this market. If it instead floods the feed with throwaway synthetic video and dents the very engagement YouTube sells to advertisers, the company will be repricing the giveaway before the next I/O.

Frequently Asked Questions

Is Gemini Omni free to use in India?

Partly. Indian creators can use Gemini Omni Flash at no cost inside YouTube Shorts and the YouTube Create app. Full access through the Gemini app and Google Flow requires a paid Google AI subscription on the Plus, Pro or Ultra tier.

How long can a Gemini Omni clip be?

Clips generated through the free Flash tier are capped at ten seconds at launch. Google describes this as a rollout limit rather than a hard ceiling of the model, so the cap could rise as the system scales.

Can it edit videos I already have?

Yes. You can upload existing footage and describe changes in plain language, such as altering a background or reimagining an action, and the model edits across multiple turns while keeping the original scene’s continuity. Audio and speech editing remain restricted while Google tests them.

Are Gemini Omni videos watermarked?

Every clip carries a SynthID digital watermark embedded in the visual signal. It is invisible to viewers but readable by Google Search and Chrome, and YouTube also adds an “AI-generated content” label to qualifying Shorts automatically.

How is this different from Google Veo?

Veo is Google’s high-end video generation model focused on output quality. Gemini Omni is built around conversational editing and multimodal inputs, letting users refine a video through ongoing instructions rather than one-shot prompts, and it ships free inside YouTube products.