Browser-based AI video generator that runs Google's Omni Flash model without your own API key, alongside twenty-one selectable video, image and music models on one credit balance.
Used this tool? Rate it
Used this tool? Rate it
Omni Flash is a browser-based AI video generator that runs Google's Omni Flash video model through a hosted API connection, turning a text prompt, a still image or a reference clip into a short video with no local install. The vendor describes it as an online AI video generator for Google's new model.
The decision is not the model — anyone can reach the same Google engine with an API key — but the packaging. The service states you can use the website without setting up your own Google API key, which removes billing setup, quota management and SDK work from the path to a finished file. You pay a credit markup and the operator's terms in exchange for skipping developer onboarding.
Despite the name, this is not a single-model tool. The workspace ships twenty-one selectable models spanning video, image and music, with a dozen competing video engines in the same prompt box; the vendor says the studio keeps Omni Flash first and also offers other video models for style testing.
One naming caveat matters before you research further: Google's own model is marketed as Gemini Omni Flash and many unrelated sites use near-identical wordmarks and domains, so ratings published under similar names do not transfer to the omniflash.net service.
The workspace exposes text-to-video, image-to-video and a first-and-last-frame mode. The frame mode changes what you can control: you supply a start frame and an end frame and the model fills the motion between them, positioned for product reveals, scene changes and short ads. The other two paths behave conventionally — describe a scene, or upload a product photo, portrait, poster or design and describe how it should move. Output is fixed to 16:9 and 9:16.
Every catalogue entry carries an explicit credit charge rather than a vague tier label. The Omni Flash model is the cheapest video option at 70 credits per generation, priced against a single 720p rate. Everything else is a multiple: Wan 2.2 at 200 credits, Grok Imagine 240, Seedance 2 Fast 250, Wan 2.5 300, Sora 2 and Veo 3.1 Lite 400 each, Kling 3.0 675, Veo 3.1 Fast 800, Sora 2 Pro 1,200 and Veo 3.1 1,600.
Image and music generation share the same wallet: Z-Image at 4 credits, GPT Image 2 at 15, Nano Banana at 20, Nano Banana 2 at 40, Nano Banana Pro at 90, and a Suno AI music entry at 60. A single Veo 3.1 render therefore costs more than twenty Nano Banana images.
Almost every model in the catalogue accepts exactly one reference image. The Omni Flash model accepts seven. That is the sharpest differentiator in the line-up and it is invisible from the marketing copy — if your work depends on feeding several product angles, character references or style boards into one shot, the default model is the only entry that supports it, and switching to Veo, Sora, Kling or Seedance drops you to a single reference. Finished clips open in a viewer with prompt copy, a first and last frame control, a recreate action, download and share, and Folders stores past work.
Rather than promising prompt engineering, the site tells users to mention the subject, action, style, camera move and length in plain language. The iteration loop, not the first prompt, is the intended workflow — the published flow ends with revising your wording and regenerating, and the viewer carries a recreate action for it.
Because per-model credit cost is published in the interface, model comparison is a budgeting exercise rather than a blind test. The vendor's method is to generate with Omni Flash first and compare others only when you want a different look, and given the spread — 70 credits against 1,600 for Veo 3.1 — validating framing on the cheap model first is the difference between one expensive render and twenty-two cheap ones.
The vendor's stated uses are product videos, social shorts, ad drafts, YouTube clips, simple explainers and quick video edits. The two aspect ratios reinforce that scope: 16:9 for landscape placements, 9:16 for vertical feeds.
Google's documentation gives a three-to-ten-second output window per generation, so these are shot-length assets, not finished sequences.
The vendor's directory listing declares the intended organisation types as small business, freelancers and individuals with browser-only delivery. There is no team or seat-based framing in the plan structure; paid tiers scale by credit volume rather than by users.
A reasonable fit if:
A poor fit if:
Delivery is web-only. The service runs in a browser with no downloads, plugins or local setup, and third-party listing data records the supported platform as browser-based cloud with nothing alongside it — no desktop or native mobile app.
The site does ship an installable web app — a prompt offers to add Omni Flash to the home screen — giving an app-like launcher on mobile without an app store listing, which matters if your organisation distributes software through managed catalogues.
Account access is federated rather than password-based, with sign-in through Google, Facebook, Yandex, LINE and Kakao — the latter three signalling deliberate reach into Russian, Japanese and Korean markets. The interface is localised into twenty-five languages, and authentication is served from a separate host domain, worth noting where network policy allow-lists domains.
Access is credit-metered. New users receive free starter credits, and more arrive through two published mechanics: a once-per-day check-in with streak tracking, and a referral programme granting credits with no purchase required from the invitee.
Three billing cycles exist — monthly, annual and one-time — each with the same three tiers. The feature ladder is what changes your rights to the output:
| Tier | Monthly | Annual | One-time | Feature ladder |
|---|---|---|---|---|
| Basic | $29 / 2,000 credits | $232 / 17,500 credits | $38 / 2,500 credits | Remove watermark |
| Advanced | $49 / 3,400 credits | $588 / 45,000 credits | $64 / 4,400 credits | Plus faster speed and commercial use |
| Professional | $199 / 14,000 credits | $1,990 / 145,000 credits | $259 / 18,500 credits | Plus priority customer support |
Two things drive most purchase decisions. Watermark removal is a paid feature at every tier, meaning unpaid output carries a watermark. More consequentially, commercial use is listed only from the Advanced tier up — Basic does not carry it at any billing cycle, so the cheapest paid entry point is not a commercial licence.
The Advanced tier is the only one with volume steps, from $49 to $1,999 monthly and $588 to $23,988 annually, with bonus credits appearing only from the second step. The one-time cycle is marketed as a one-time payment with no auto-renewal, and card payment runs through a single live provider; other listed methods are unavailable or coming soon.
Translating credits into output needs the model table, not the plan table. At 70 credits per generation the Basic monthly allocation funds roughly 28 renders of the default model, but only one Veo 3.1 render. Vendor-supplied directory figures quote plan capacity using a 30-credit-per-video divisor matching no model in the catalogue, so "up to N videos" claims should be checked against per-model costs.
The most useful comparison is not against other websites but against the two other ways to reach the same engine, since the output is identical in all three.
Going direct to Google. The engine is available through Google's own developer surfaces, where independent editorial coverage puts list pricing at roughly $0.10 per second of 720p output. Direct access adds API key provisioning and billing setup but removes the reseller's terms, refund ceiling and watermark tier entirely.
Staying inside the same studio. Because twenty-one models share one wallet, the nearest alternatives are already in the product. Veo 3.1, Sora 2, Kling 3.0, Seedance 2 and Wan 2.7 sit in the same selector at published credit costs, so switching engines costs no new subscription — only a different credit charge and a single reference image.
Independent catalogues. Third-party directories place this service in the AI video generator category alongside browser-based peers such as Runway, Kling AI, VEED and HeyGen, and one catalogue that links to this operator lists Ray 3.2, Veo 4, Wan 2.7 and Seedance 2.5 as nearest similar tools. That catalogue shows no accumulated user reviews, so peer comparison rests on feature and price parity rather than user experience.
Discretionary refunds require two conditions at once: a time window and a consumed-credit ceiling. Monthly subscriptions and one-time credit packs allow a refund within three days and only if no more than 30 credits have been consumed; annual subscriptions allow five days and 50 credits. Because the default model costs 70 credits per generation, a single successful render already puts an account past every one of those ceilings, so the discretionary refund covers evaluating the interface rather than the output. Refunds are also refused outside the window and for chargebacks raised without contacting support first. Failed generations are treated separately: if credits were debited and not auto-refunded, the vendor states it will make it right at no cost, with a support response target of five business days.
The terms grant ownership of generated output for personal or commercial use but attach two commercially significant limits. They note that AI-generated output may not always be eligible for copyright protection, and separately prohibit reselling, sublicensing, scraping or otherwise commercialising the raw output — so building a downstream service on the generations, as distinct from using them in your own marketing, falls outside the granted rights. Commercial use also appears only from the Advanced tier up, so your licence depends on the plan as well as the terms.
The service is provided on an as-is and as-available basis without warranties, and the vendor does not warrant that output will meet expectations — pointed where each unsatisfactory render is a non-refundable charge. Access may be suspended at any time, and because generation is routed through a third-party provider the vendor disclaims responsibility for those providers' availability, so engine-side outages fall outside the guarantee. Use is restricted to those aged 18 or over, and the terms prohibit pornographic or gratuitously violent output and content depicting real identifiable people without consent.
Prompts, uploaded reference files, generation parameters, outputs, credit balances, transaction history and timestamps are all stored. Uploads and prompts are passed to AI inference providers to fulfil the request, so reference material leaves the vendor's own systems by design — relevant if you upload unreleased product imagery. Generation history is retained with a default minimum of 30 days and server logs for up to 90 days, and users may have the right to access, correct, export or delete their personal data by emailing support.
Some limits belong to Google's model rather than this front end, but they bound what any reseller can deliver. Google's documentation gives a three-to-ten-second single-generation window, and the model card names edit consistency, complex motion and accurate text rendering as open weaknesses. Independent editorial coverage adds that 1080p and 4K are upscaled rather than natively generated, and that editing or extending uploaded video is unavailable in the EEA, Switzerland and the UK. Here the default model is offered only against a 720p rate.
It depends on the model. The namesake model is the cheapest video option at 70 credits per generation. Others range from 200 credits for Wan 2.2 to 1,600 for Veo 3.1, with Sora 2 Pro at 1,200 and Kling 3.0 at 675. Image models run from 4 credits to 90.
Watermark removal is listed as a paid plan feature at every tier, including the cheapest, so output generated without a paid plan carries a watermark. Separately, the underlying Google model applies SynthID provenance marking as a safety measure.
Rarely. Discretionary refunds require both a short window — three days for monthly plans and one-time packs, five for annual — and a consumed-credit ceiling of 30 or 50 credits. Since one render of the default model costs 70 credits, generating even once normally forfeits eligibility. Failed generations where credits were debited are handled separately and made right at no cost.
Seven for the Omni Flash model, the highest in the catalogue by a wide margin. Every other video and image model in the studio accepts a single reference image, and Z-Image accepts none at all.
Not a native one. The service is browser-based and offers to install itself as a home-screen web app, so it behaves like an app on a phone without being distributed through an app store.
