Creating AI images for free: what the free tiers really offer

This article is an overview based on vendor information, documentation and publicly available reporting, not a scored test of our own. Prices and features change quickly; the vendor's own terms always take precedence. We only score what we have used ourselves for weeks: those tools are listed under AI tools compared.

Free almost always means: with an account and with a quota

Search for ways to create AI images for free and you will find dozens of offers, almost all built on the same pattern: sign in with an email address or an account from one of the big vendors, then a quota that refills daily or monthly. Genuine use without signing in has almost disappeared, and where it is offered the quality is usually a generation behind.

The quotas themselves move. Most vendors do not publish a fixed number and adjust the limit to demand; what applied in spring may be halved by autumn. The figures in this article are therefore orders of magnitude, as of September 2026.

Three other questions matter more than the number: who can see the images you create? May I use them commercially? And what happens to my prompts? On these three points the free tiers differ more than on image quality.

The assistants: Gemini, ChatGPT and Copilot

The most convenient route goes through an assistant many people already use. In the free Gemini app, Google's image model "Nano Banana" reportedly produces up to about 20 images a day; the stronger Pro variant requires the Google AI Plus subscription, which costs 4.99 euros a month in Germany. Every image carries an invisible SynthID watermark identifying it as AI-generated. More on the assistant itself is in our review of Gemini.

ChatGPT also creates images on the free plan, but only a few a day. OpenAI does not publish a fixed number; users report two to three images within 24 hours. Since the end of August 2026 ChatGPT has also been showing ads in Germany on the free plan and the low-cost Go plan.

Microsoft is currently the most generous. Bing Image Creator, also built into Copilot, allows 15 fast images a day according to Microsoft's help page, after which it continues more slowly, up to a ceiling of 200 prompts in 24 hours. The terms of use expressly permit any lawful use, including commercial use. A Microsoft account is required.

The specialists: Firefly, Ideogram, Canva and Leonardo

Adobe Firefly, ranked first in our AI image generators category, offers free daily generations on a selection of models; Adobe no longer states a number. Its real advantage lies elsewhere: Adobe trains its own models on licensed material and markets them as safe for commercial use. For images that will end up on a company website, that is the best starting position among the free options. The Standard plan with 2,000 credits costs around 10 dollars a month.

Ideogram is the first port of call when text has to appear in the image, on posters, invitations or social media tiles. The free tier offers only a few slow generations, and the images are publicly visible in the service's gallery. For a product visual ahead of a launch, that rules it out.

Canva reportedly gives users without a subscription 20 AI uses a month, shared across all AI features including image generation. Leonardo reportedly hands out 150 tokens a day, enough for a few dozen images; free images are public here too.

Midjourney: no free access

The question comes up often and the answer is short: Midjourney has had neither a free tier nor a trial since 2023. The cheapest plan costs 10 dollars a month, or 8 dollars on annual billing. Offers promising "free Midjourney" are either third-party services running other models or simply advertising.

Worth knowing even on a paid plan: images from the smaller tiers appear in the public gallery. Private mode only comes with the more expensive plans. Anyone developing client visuals with Midjourney therefore needs more than the cheapest tier.

Unlimited and private: Stable Diffusion on your own computer

The only route without a quota, without an account and without any image leaving your own machine is an open model such as Stable Diffusion or Flux, run through an interface such as ComfyUI or Forge. The software costs nothing; the graphics card pays the bill.

As a rule of thumb for video memory: Stable Diffusion 1.5 runs from about 4 GB, SDXL needs at least 8 GB and runs smoothly with 12 GB, and Flux.1 at full precision needs 12 to 16 GB; compressed versions manage with less. Without a suitable card, a rented cloud machine helps, but then it is no longer free.

The price is the learning curve. In our test Stable Diffusion scored only 2.9 out of 5 for ease of use, but the best marks in the field for value for money and data protection. If you regularly need many images and do not mind a week of learning, it is the cheapest option in the long run. How to run language models locally in a similar way is covered in our article on local AI.

Rights, visibility, labelling

Clarify three things before a free image goes public. First, visibility: on Ideogram, Leonardo and Midjourney, images from the cheaper tiers are public, together with the prompt that produced them. Second, the terms of use: some vendors allow commercial use even on the free plan, others only with a subscription, and the rule usually applies to the moment of creation.

Third, copyright. An image created without any significant human creative input is, according to the prevailing view in Germany, not a protected work, because the law requires a personal intellectual creation. You may use it, but you can hardly prevent others from doing the same. For a logo or a central brand visual, that is a weak foundation.

In addition, the transparency obligation of the AI Act has applied since 2 August 2026: anyone publishing realistic images of real people, places or events must label them as artificially generated. There is no general obligation to mark every illustration. The checklist for business use is in our article on GDPR and the AI Act.

Which route for whom

For occasional private images, Gemini or Copilot are entirely sufficient, both without an extra account if you already have one with Google or Microsoft. For a small business's social media posts, Firefly is the safest start; if the daily quota is not enough, the Standard plan is the cheapest clean solution.

If you need text in the image, try Ideogram; if you want many images and full control, set up Stable Diffusion. And if you are after a distinctive visual style, you will sooner or later end up on a paid Midjourney plan. How far free tiers go in other areas is covered in our article on free AI tools; how to turn an image into a moving clip, in the guide Image to video with AI.

In short: with 15 fast images a day, Copilot is the most generous free tier, Gemini the most convenient, Firefly the safest in legal terms. Ideogram for text in images, Stable Diffusion for unlimited and private images on your own hardware. Midjourney is subscription-only. With free images, check visibility, commercial use and labelling first.