A 15-second AI video clip now costs about $1.05 at 720p

A full 15-second AI video at 720p now costs about $1.05 on Google's newest Omni model. On November 5, 2026 Google DeepMind and Google Cloud announced Gemini Omni 1.2 Flash as generally available on the Gemini API and Gemini for business. The jump from Omni 1.1 is not a research teaser. It is a production model with a longer native clip length, a lower per-second price, multi-turn conversational editing, stronger reference adherence, SynthID watermarking, and Provisioned Throughput for enterprise traffic.

That combination matters for anyone who already ships product spins, try-on clips, catalog videos or short social ads at volume. Fifteen seconds is long enough for a complete social spot or a short explainer without stitching. A 30% price cut versus 1.1 Flash cuts the bill on every regeneration. Multi-turn edits mean you can swap a background or product without paying for a full new take every time. Below is what shipped, what it costs, where it runs, and what Google still says needs work.

What shipped on November 5

Google's primary post frames four product changes at once. First, native generation and editing now reach 15 seconds, up from 10 seconds on Omni 1.1. Duration is selectable from 3 to 15 seconds in one-second steps, so a three-second bumper and a fifteen-second ad spot sit on the same model. Second, Google cut the price by 30% versus Omni 1.1 Flash, to $0.07 per second of 720p video. Third, multi-turn conversational editing lets you refine the same clip: swap a product or background, change lighting or weather, or switch aspect ratio without starting over. Fourth, reference adherence is stronger across character, product and style images, short motion clips up to three seconds, and audio cues.

Google Cloud's companion post repeats the same numbers for enterprise buyers and adds the production posture: Provisioned Throughput is available, outputs carry an imperceptible SynthID watermark, and the model is backed by Google Cloud infrastructure and security. Audio references are gated behind allowlisting. That last detail matters if your workflow depends on voice-conditioned characters on day one.

Secondary coverage from MarkTechPost summarizes the same release for developers: no open weights, no local deploy, hosted only on Gemini API and Google Cloud, with Google still flagging fast sports action, rapid script changes and complex physics as weak spots.

Price, resolution and the draft-then-upscale loop

The headline rate is $0.07 per second of 720p. A 10-second 720p clip is about $0.70. A 15-second 720p clip is about $1.05. Google's Gemini API pricing page explains the token math behind that number: about 5,792 output tokens per second of 720p at $12.00 per million tokens, which lands near $0.07 per second. The same page lists approximate effective rates of about $0.023 per second at 360p, $0.104 per second at 1080p with upsampling, and $0.209 per second at 4K with upsampling.

The intended workflow is not "always render 4K first." Google pushes a draft loop: generate fast at 360p or 720p, fix details with follow-up edits, then upscale the keeper to 1080p or 4K. That is how the 30% cut compounds. You spend less per failed take, and you only pay the higher resolution rates once the creative is locked.

Omni 1.1 Flash is already on a sunset path in the API docs. Effective December 3, 2026, requests to `gemini-omni-1.1-flash` automatically route to `gemini-omni-1.2-flash` and bill at 1.2 rates. Teams still hardcoding the old model name should plan the cutover rather than discover it as a surprise invoice change.

Longer clips without giving up control

Ten seconds was already useful for hooks and product turns. Fifteen seconds is the length of many complete social ads, CTV bumpers and short explainers. Google's demos stress unbroken camera moves, first-to-last frame consistency, and motion transfer across turns: for example, generating a rippling octopus and then replacing it with a dancer whose gown mirrors the same fluid motion. Virtual try-on sequences can keep a model's identity while rotating through several outfits across a 15-second spin.

Consistency is the other half of the story. Google claims stronger hold on faces, garments, packaging typography and brand colors across camera angles, plus less morphing when interpolating between keyframes or extending a clip. That is the feature catalog teams care about when the alternative is re-rolling until a logo stops melting. Independent third-party scores for 1.2 versus 1.1 are not in Google's posts. Customer quotes on the Cloud blog (Artlist, FLORA, invideo, Krea, Picsart, Photoroom, QuickFrame, Runway) are endorsements, not public benchmarks.

Speech also got an upgrade pass: more fluid dialogue, clearer lip-sync, and distinct multi-speaker conditioning. That shows up in Flow scene direction, Google Vids explainers with avatars and narration, and multi-character dialogue. Controllable camera work is pitched the same way: sweeping FPV flythroughs, zoom-outs and orbits with fewer unwanted jump cuts.

Where you can actually use it

Availability is broad for a Google model launch:

• Developers: Gemini API paid tier, with prototyping in Google AI Studio and docs plus pricing linked from the launch post.
• Businesses: Google Cloud console, Gemini for business, and API embedding into apps and agents. Provisioned Throughput for predictable capacity.
• Creators and Workspace users: Flow for scene direction and multi-turn edits; Google Vids can generate up to three 15-second videos per month at no additional cost with a Google account or Workspace subscription; the Gemini app is rolling 15-second video to Google AI Ultra subscribers; Google Ads can scale video campaigns.

That spread is the strategic point. Google is not only selling seconds of video generation. It is putting the same model under consumer Vids quotas, creator Flow tools, Ads inventory and Cloud SLAs. Photoroom's CEO told Google that 15-second length was a top user request for its Video Ad Maker and Product Animator. QuickFrame AI called out 15-second 1080p generations as enough for a complete CTV ad in one pass. Those are volume surfaces, not demo reels.

What this means if you make product video

If you already pay for hosted AI video, Omni 1.2 Flash changes three practical knobs: length, price and edit cost. A catalog team can aim for complete 12 to 15 second spots instead of stitching 8 to 10 second fragments. A social team can keep more variants because a follow-up edit is cheaper than a full regenerate. An enterprise buyer can ask for Provisioned Throughput and expect SynthID on outputs for provenance.

If you are comparing vendors, treat Google's claims as product claims until you run your own evals. Google openly says fast sports, rapid script changes and complex physics still need work. Audio references need allowlisting. There are no open weights. There is no local deployment path. For many sellers and agencies that is fine. For teams that need offline or open-weight pipelines, Omni is not that product.

Also watch the December 3 routing change. Any automation still pointed at `gemini-omni-1.1-flash` will land on 1.2 pricing and behavior whether or not you renamed the model string. That is good if you want the upgrade. It is a surprise if your cost models still assume 1.1 rates.

Competitive context in plain terms

The AI video market has been a race on seconds, adherence and dollars per clip. OpenAI, Runway, Luma, Kling and others have all pushed longer generations and cheaper drafts in 2025 and 2026. Google's move is conservative in one way and aggressive in another. Conservative: a five-second length bump and a 30% price cut, not a sudden leap to minute-long narrative film. Aggressive: general availability across API and Cloud on the same day as consumer Vids quotas and Ads paths, with watermarking and Provisioned Throughput already attached.

For brands that live inside Google Workspace and Ads, the switching cost falls. For brands that standardized on another video stack, Omni 1.2 Flash is another vendor to A/B on product fidelity, text readability on packaging, and how often a logo survives a camera orbit. The $1.05 figure for a 15-second 720p clip is a useful yardstick only if your quality bar matches Google's demos. Failed takes still cost money. Multi-turn editing is how Google tries to keep those failures short.

What to do next

Developers should open the Gemini API docs, price a 15-second 720p and 1080p path against current spend, and test multi-turn edits on one real SKU with packaging text. Business teams on Cloud should check Provisioned Throughput and SynthID requirements with legal and brand before scaling. Workspace users can burn the three free Vids generations to see whether 15-second clips clear their bar before buying more capacity.

If your bottleneck is physics-heavy action or sports, wait for Google's next pass or keep a second vendor in the loop. If your bottleneck is volume product video with tight brand references, Omni 1.2 Flash is the release Google built for that job: longer, cheaper per second, editable in conversation, and marked GA.

The bottom line

Gemini Omni 1.2 Flash turns a 10-second flash model into a 15-second production option at $0.07 per second of 720p, with multi-turn edits, stronger reference hold, SynthID and Cloud-grade availability. The surprising number is not the marketing slogan. It is that a complete 15-second 720p clip now sits near a dollar, and you can revise that clip without regenerating the whole thing. Google still owns the weak spots around sports, physics and allowlisted audio. For catalog, try-on and social ad factories, the November 5 GA launch is the moment to re-run cost models and adherence tests against whatever you used last month.

Sources: Google DeepMind / blog.google, Google Cloud Blog, Gemini API pricing, Google Vids announcement, MarkTechPost.