A browser-based generative suite that routes one prompt to Sora, Veo, Kling, Flux and other third-party image and video models, with credit-based plans and built-in editing.
Used this tool? Rate it
Used this tool? Rate it
DaVinci AI is a browser-based generative media suite that routes a single prompt box to third-party image and video models instead of running one proprietary model of its own. It sells access rather than a house style. The homepage names Sora, Veo, Nano Banana, Kling, Seedance, Seedream and Flux among its featured engines. The operator is HubX Yazılım Hizmetleri Anonim Sirketi, registered at a Konak address in Izmir, Türkiye.
The vendor states its scale on the homepage as more than 100 million creators worldwide. It is an unaudited self-published figure with no methodology attached, so it should not carry weight in a purchase decision.
One naming caution: this product is unrelated to the Blackmagic Design video editing application of a similar name and to any legacy language model that once shared the word.
As an AI image generator, the capability set splits into generation and post-generation editing, and editing is where it tries to earn a place in a working routine.
The mobile apps are advertised as carrying more than fifty models and covering image, video and audio without a desktop, which is the whole promise of a multi-model AI platform compressed into a phone. Independent testing reports that a new account can run its first image without entering card details, which lowers the cost of finding out whether the catalogue suits your material.
That breadth carries a cost. Reviewers report that the size of the catalogue itself creates a learning curve, because matching a model to a job takes trial and error.
Four editing operations are named on the product pages, and each one targets a specific failure mode of a finished generation rather than a fresh prompt.
The image generator page advertises instant access to over 15 industry-leading models in one workspace, naming Nano Banana, Seedream, GPT Image 2 and Kling among them. That is the shape of the multi-model pitch: the first choice is which engine to spend on, not which style preset. Independent testing adds a tier boundary: running several generations at once to compare results side by side is reserved for higher-tier plans, so cheap seats compare models one run at a time.
The published flow runs in three steps: open the AI image and video tool and choose the image generator, select the model and write the prompt, then generate and refine the result. The order carries a cost implication the step list omits.
Credits are spent only on generation, never on downloading finished work, and the cost per run varies with the model and the output type. Two consequences follow. Exporting the same asset in different formats costs nothing, so there is no reason to hoard downloads. And the price of a run is not flat, so the AI video generator draws the balance down far faster than an image attempt.
Sequencing has a price: settling a composition on cheap image runs before committing to expensive video runs separates a productive month from an exhausted entry-tier allowance.
The suite runs in a browser on desktop, tablet and mobile, and ships as native apps on both stores. The mobile apps are not a thin wrapper around the same account experience, and three listing details change what a mobile buyer actually gets.
| Surface | Rating and sample | Notable listing detail |
|---|---|---|
| Web workspace | Not store-rated | Full generation and editing surface |
| iOS app | About 4.54 from 75,328 ratings | 12+ content rating; iOS 16.0 minimum; five interface languages |
| Android app | 3.5, counted as 186K reviews at the top of the page and 181K in the ratings block | Over ten million downloads |
Apple's own app metadata reports about 4.54 across 75,328 ratings, lists the content advisory rating as 12+, and sets iOS 16.0 as the minimum operating system. The Android listing for the HubX-published app sits at 3.5 against more than ten million downloads, but its review count is not stated consistently: the top of the page reads 186K reviews while the ratings block further down the same page reads 181K. Treat the Android sample as roughly 180,000 reviews rather than one exact figure; side by side, the family scores about a point higher on iOS than on Android.
Both diverge sharply from the store rating the vendor advertises on its own homepage, which is presented as a 4.9 across more than 120,000 reviews. That figure matches neither listing, so the two published store scores are the ones to plan around.
Language coverage is narrower on mobile than the web suggests. The iOS interface itself ships in five languages: English, Japanese, Korean, Portuguese and Spanish.
Every advertised monthly figure on the plan ladder is an annually billed rate, so the headline number is not what a month-to-month buyer pays.
| Plan | Advertised rate | Monthly credits | Image ceiling | Video ceiling |
|---|---|---|---|---|
| Max | $39.99 / month, billed annually | 5,000 | 2,500 | 125 |
| Ultimate | $99.99 / month, billed annually | 15,000 | 7,500 | 375 |
| Creator | $199.99 / month, billed annually | 40,000 | 20,000 | 1,000 |
Max sits at 39.99 dollars a month billed annually and carries 5000 credits, up to 2500 images and up to 125 videos each month. Ultimate raises that to 99.99 dollars with 15000 credits, up to 7500 images and up to 375 videos. Creator, flagged as a limited offer, reaches 199.99 dollars with 40000 credits, up to 20000 images and up to 1000 videos.
The image and video ceilings are not independent budgets but two expressions of the same credit pool, which is why the ratio holds across all three tiers at twenty images or one video per hundred credits.Pick a tier by the video ceiling, since that binds first for anyone producing motion.
Credits refresh to the plan allowance at the start of each billing period rather than accumulating. An underused month is lost. A trial or introductory period rolls into the standard price automatically unless it is cancelled before renewal, which is the single most common source of unexpected charges on subscription products of this shape. Price rises come with at least thirty days of notice and land only at the next renewal, so an increase can always be refused by cancelling inside that window.
One stale artefact matters when comparing quotes. A legacy pricing route built on a daily credit allowance still surfaces in search caches, but a direct request to it now returns a not-found response. Figures from that cached page do not describe the current offer.
A software directory's review article places it against Midjourney, Krea, Higgsfield and Freepik AI, and the split is breadth against specialisation. The arithmetic favouring consolidation is simple: separate subscriptions to a leading video model, a leading image model and an upscaler stack up quickly against one tier here. The case for a specialist applies whenever one model does almost all of your work, since the rest of the catalogue goes unopened.
The single largest gap between the marketing surface and the binding document is the declared purpose of the product. Marketing addresses campaigns, brands and creative teams; the licence restricts the service to personal, entertainment and creative use and rules out professional and commercial contexts.
Commercial use is permitted while a subscription stays active, ownership sits with the creator, and responsibility for copyright, trademark and identity compliance sits there too. The permission is conditional: it tracks an active subscription rather than vesting at the moment of creation.
Exclusivity is expressly disclaimed. Outputs may not be unique, other users may receive the same or similar results, and no guarantee is offered that any output qualifies for intellectual property protection. For a logo or brand mark meant to be defensible, that disclaimer is disqualifying.
Two versions of the training promise sit on the same homepage, and they do not carry the same boundary. The support answer states flatly that inputs and outputs are not used to train the models and that outside model providers are held to the same rule. A marketing block on the same page softens that into a promise that data is never used for training without permission. An unconditional statement and a consent-gated one are different commitments, and only the weaker survives if consent is collected by default.
Collection continues regardless. Advertising identifiers on both mobile platforms are collected, and location data is gathered to localise and personalise content. Serving personalised advertising is listed among the declared purposes for the information collected.
The Android data safety panel states that data is not encrypted in transit and that data cannot be deleted through the store request route. Both come from the developer, and for anyone uploading identifiable photographs this is the most consequential disclosure on the page.
Apple's privacy panel for the same app declares that Contact Info and Identifiers may be used to track you across apps and websites owned by other companies. That is a different exposure from the generation data handled by outside model providers: it follows the account outside the product entirely. Anyone bound by an employer or client rule on cross-company tracking should read this before installing the mobile app rather than the web version.
Small inconsistencies recur across the vendor's pages. No attribution to the DaVinci name, logo or link is required, though that answer names Basic, Pro and Creator plans while the ladder above it sells Max, Ultimate and Creator. A consumer review platform profile for the domain is a claimed business profile carrying 4.5 across roughly 5,400 reviews, so on that channel the vendor scores well above its Android listing and in line with its iOS one.
No. This is a browser-based generative image and video platform operated by HubX from Izmir, Türkiye, with no relationship to the Blackmagic Design editing application or to any older language model of a similar name.
Independent testing reports that a new account can run its first image without entering card details. Video generation stays locked behind a paid tier for anyone on the free plan, so the free entry point evaluates the image side only.
No. The balance resets to the plan allowance each cycle instead of accumulating, so a quiet month buys nothing toward a busy one.
The vendor says yes for as long as a subscription is running, with ownership and legal responsibility both resting on the creator. Weigh that against the licence clause confining the service to personal, entertainment and creative purposes, and against the exclusivity disclaimer covered under limitations.
The vendor's answer is no, but the homepage carries that assurance twice in two different strengths, and only the weaker of the two is qualified by consent. The limitations section sets both versions side by side, because the difference decides whether the protection is unconditional or something a checkbox can undo.
Some image features run on outside providers, and image data is handed to the relevant provider to process each request in real time. Advertising and tracking identifiers are a separate matter from generation data, and the limitations section sets out what each mobile platform declares on that front.
Removing the app from a device settles nothing; the subscription lives with the store or account that bought it and has to be ended there. Refunds are the harder half of that answer, and the section above sets out why the timing of a cancellation matters more than the reason for it.
Users in the European Union, the United Kingdom and Brazil are named as holding rights to access, correction, deletion, restriction, portability and objection. Serving personalised advertising is listed among the declared purposes for the information collected, the processing most users will want to object to first.
