Craiyon turns typed descriptions into images in the browser across six style presets, with a free watermarked tier and paid plans that meter editing credits rather than generation.
Used this tool? Rate it
Used this tool? Rate it
Craiyon is a browser-based text-to-image generator: you type a description, pick a style, and the service returns generated images in seconds. As a free AI image generator it needs no download, and as an AI art generator it leans on style presets rather than on parameter dials. No local model and no signup wall stand in front of the first generation.
The product launched as DALL·E mini in 2021 and took the Craiyon name in 2022. That history explains why the tool still turns up under the old name in bookmarks and forum threads. Independent reporting published in July 2022 recorded that OpenAI asked the creator to change the DALL·E mini name, and, citing a model-hosting platform, put the model's output in 2022 at up to 50,000 images per day. Those figures describe the 2022 model and the traffic it carried then, not the generation of models the site serves today.
The service is operated by Craiyon LLC, a company organised in Delaware, and the terms require users to be at least the age of majority where they live. Two things about the current product decide most adoption questions: the free tier is genuinely usable but visibly marked, and the paid tiers are priced against editing volume rather than against generation volume. Read the plan table with that second point in mind, because it is the opposite of how most image services meter their subscriptions.
Six style presets shape the output before you touch the wording of the prompt. Artistic covers digital painting and creative texture, Photo aims at realistic photographic results, Anime produces Japanese-animation-inspired visuals, Illustration handles drawn and digital art styles, Vector yields clean shapes suited to scalable graphics, and Raw applies no style bias at all.
The credit model is the part worth understanding before you compare plans. Generating from a text prompt alone consumes no credits at all; credits are spent only when an image goes in as input and gets retouched. That inverts the usual metering assumption: text-to-image volume is not what you are buying, editing capacity is.
A free account returns base-quality images carrying a watermark and can be made to wait during busy periods. Paid accounts reach the highest-quality models, unlock every aspect ratio, download at higher quality, drop the watermark, and skip the queue. The quality gap is a model gap rather than a resolution setting, so it is not something post-processing recovers: upscaling a base-quality result does not reconstruct the detail the higher-quality model would have produced in the first place.
The queue behaviour deserves a note of its own. Free generation can return a busy message at peak times, and the documented response is simply to try again shortly rather than to change any setting. Subscribers are placed ahead of that queue, which is why plan choice affects turnaround as much as it affects output quality.
The shortest productive loop is prompt, style, exclude, refine — in that order, because each step narrows a different failure mode.
Concept art, social visuals and reference images are where the free tier holds up best. In each of those cases the image is an input to a later decision — a mood direction, a placeholder, a thumbnail test — and a watermark on a draft costs nothing. The public prompt corpus makes this faster than it sounds: reading a working prompt for a style you cannot reproduce is usually quicker than iterating blind.
The editing path serves a different job. Because retouching accepts an existing image as input, the tool can be used to adjust an asset you already have rather than to originate one, and that is the work the monthly credit allowance is sized for. A practical consequence is that teams generating heavily but editing rarely may sit comfortably on the smallest paid tier, while teams that treat every output as a starting point for revision will exhaust credits far faster than their image count suggests.
Multilingual prompting widens this further for non-English workflows. A brief written in the language your team actually works in can go straight into the prompt field without a translation step, and the descriptive detail you would add anyway carries more weight in the result than the language switch does.
The fit question turns on whether a watermark and a queue are acceptable in your output path. If the image is a draft, a reference or a personal piece, they are not obstacles. If the image ships directly into a client deliverable, the free tier is a trial rather than a workflow.
The free tier is capped in both volume and functionality rather than being an unlimited sandbox, so evaluating it as "unlimited free generation" will misprice the decision. Budget the trial as a sample of quality, not as ongoing capacity.
There is also a governance dimension that decides fit independently of budget. Because a subscription is bound to one individual and may not be shared across people or accounts, a shared team login is not a supported way to spread the cost, and an agency serving several clients has to plan seats rather than a single account.
Everything runs in the browser, and that is currently the only supported entry point. No official mobile application exists yet; the vendor describes one as forthcoming, which means any app carrying the name in a store today is not the vendor's own.
When images refuse to appear, the documented sequence is to switch to Chrome or an incognito window, clear the cache, turn off any VPN, and only then try another device. The ordering is informative: the common causes are browser state and VPN routing rather than the account or the plan. If a corporate network routes traffic through a VPN by default, that is worth testing before raising a support ticket.
Browser-only delivery also sets the practical ceiling on how the tool integrates. Assets leave through the download path and the account workspace, so anything downstream — a design file, a CMS, a slide deck — receives them as ordinary image files rather than through a connector.
Craiyon pricing puts three paid tiers above the free account, with yearly totals billed at $120 for Supporter, $240 for Professional and $840 for Ultra. The yearly option is advertised as 17% off the monthly rate.
| Plan | Billed yearly | Monthly edit credits | Image visibility | Queue |
|---|---|---|---|---|
| Free | — | none | public | may wait at peak |
| Supporter | $120 | 100 | public | priority |
| Professional | $240 | 220 | private available | priority |
| Ultra | $840 | 900 | private available | priority |
Edit credits are the real differentiator between the tiers: 100 a month on Supporter, 220 on Professional and 900 on Ultra. Since text-only generation is uncharged, the tier you need is set by how much retouching you do, not by how many images you make. Unused monthly credits do not roll over on any plan, and buying a year in advance does not enlarge the monthly allowance — the yearly saving is on price, not on capacity. Top-up credit packs can be purchased separately when a month runs short.
The pricing page carries a monthly/yearly switch, and the $120, $240 and $840 figures are the amounts charged when the yearly side is selected. Switching to monthly billing raises the annual outlay by the advertised 17% but leaves the credit allowance untouched, since yearly plans receive the same monthly credit allowance as monthly plans.
Subscriptions and one-time credit purchases alike are final, and a partially used billing period is not refunded. Cancelling stops future charges and leaves access running to the end of the paid term, so the decision to cancel is best made before a renewal date rather than after it.
Two conditions push a workload off Craiyon and toward alternatives regardless of price. The first is programmatic access: without a public API, any pipeline that has to generate images from a server or a scheduled job cannot be built here at all. The second is a hard requirement for private generation at the entry price point, since the cheapest paid tier still publishes what it makes. A third, softer condition is attribution: if a publishing workflow cannot reliably attach credit text to every image, staying on the free tier for commercial output is not viable and the comparison becomes paid-Craiyon against paid alternatives rather than free-Craiyon against them.
Attribution is the boundary that catches free users out most often. Commercial use is allowed on any account type, but a non-subscriber must credit Craiyon in text accompanying the image; placing the logo in a corner counts only where accompanying text is genuinely impossible. That is a stricter condition than "free for commercial use" suggests, and it applies per published image.
Supporter keeps generated images public, so private generation starts at Professional rather than at the first paid step. Teams that treat unreleased campaign visuals as confidential should read that as a plan requirement, not a preference.
The licence you grant on prompts and images is broader than most people expect from a free drawing toy. Under the terms, Craiyon receives a perpetual, worldwide, non-exclusive, sublicensable, royalty-free and irrevocable copyright licence to reproduce, adapt, display, perform, sublicense and distribute both the prompts you enter and the images produced at your direction, and that licence survives termination of the agreement. You are not signing your images away exclusively, but you are not withdrawing them either.
Advertising and analytics sit inside the free experience by design. The privacy policy names Google Analytics and states the site may use analytic or marketing services including Google AdSense, Taboola or RevContent, with consent treated as given by use. It also states the site does not sell, rent or disclose collected information to outside parties except in named cases such as affiliated service providers, legal demands and business transfers.
Yes, generation is available without payment, but the free account is limited in both the number of images and the functionality available, and output carries a watermark at base quality.
The generator runs in the browser and the first images do not require a purchase. Paid features — model quality, aspect ratios, watermark removal and priority — are what a subscription buys.
Commercial use is permitted on any account type. Without a subscription you must credit Craiyon in text alongside the image, and a corner logo is acceptable only where accompanying text is impossible.
Credits meter editing. Generating an image from text alone uses none; feeding an image back in and retouching it with the highest-quality models is what draws down the monthly allowance.
No. Unused monthly subscription credits expire and refresh on the billing date, on monthly and yearly plans alike, and a yearly plan receives the same monthly allowance as a monthly one.
Not on the free or Supporter tiers, where generated images are public. Private generation begins at the Professional tier.
Neither at present. The vendor states a public API is planned and a mobile app is coming, so any app currently listed under the name in a store is not the official one.
Yes. Prompts can be entered in any language, and adding descriptive detail matters more to the result than the language itself.
