Anyone with a phone can now create images with AI. You type a short description and a picture appears, or you upload a photo of yourself and ask for changes. Many people in the Gulf search for ways to turn their own photo into a cartoon or place themselves somewhere else. This page explains calmly how an AI image generator works and introduces a few, without ranking them. It then looks at what matters before you upload your face or your child's, and at what Gulf laws say about other people's images. Each tool's facts come from the company's own pages, as of 31 October 2026.

The short answer

  • What are they? Tools in the family of generative AI that create a new image from a written description, or edit an image you upload.
  • Are they free? Some have limited free use, such as ChatGPT, Gemini and Leonardo.Ai. Others need a subscription, such as Midjourney on its website.
  • Before you upload your photo: find out whether images are used to improve the service and whether your images are visible to the public, and check the settings.
  • Other people's images: the UAE and Saudi Arabia have laws that punish invading someone's privacy through their images, and altering a photo to defame its subject is a crime in the UAE. Always ask permission.

How does an AI image generator work?

These tools are built on models trained on huge numbers of images and the text that goes with them, so they have learned how words relate to shapes. When you write "a cup of Arabic coffee on a wooden table at sunset", the model creates a new picture that matches the description. It does not search for an existing one.

They work in two main ways:

  1. Creating an image with AI from scratch. You write the description and get one or more images to choose from.
  2. Editing a photo you upload. Changing the background, adding an object, or turning your photo into a drawing. Privacy matters more here, because you are sharing a real photo.

The result is not a photograph of something that happened. It can look very convincing, but the model made it, and it may contain mistakes in hands, text or small details.

Well-known tools: ChatGPT, Gemini, Leonardo AI and Midjourney

The table sums up what each company says about its tool, as we checked it on 31 October 2026. We do not rank these tools or recommend one, and there are many others.

ToolMade byFree use, according to the companyPrivacy note
ChatGPTOpenAIThe free plan can create images, within usage limitsYou can stop your chats being used for training
GeminiGoogleImage creation is available once signed inEditing images is not available to under-18s
Leonardo.AiLeonardo.Ai, owned by Canva150 "fast tokens" a day on the free planFree-plan creations are public
MidjourneyMidjourneyNo free trial on the website; a limited trial in the niji · journey appPrivate mode is only on the two top plans

ChatGPT, from OpenAI. OpenAI's free-plan page says free users can create images in ChatGPT, and that paid plans have higher usage limits (OpenAI). The company's list of supported countries includes all six GCC states, as we explain in our guide What is ChatGPT? How to use it in Arabic, safely.

Gemini, from Google. Google's help page says you must be signed in to create images, must be 13 or over (or the applicable age in your country) to generate images, and must be 18 or over to generate and edit images (Google). Google lists Arabic among Gemini's languages and all six GCC states among the countries where the web app is available (Google).

Leonardo.Ai (Leonardo AI). The Australian company Canva acquired it in July 2024 (TechCrunch, 29 Jul 2024; Bloomberg, 30 Jul 2024). Its pricing page shows a free plan with 150 "Fast Tokens" a day and describes that plan's creations as "Public"; paid plans start at $12 a month (Leonardo.Ai).

Midjourney. The company's documentation says: "No free trial is currently available in Discord or the midjourney.com website." It adds that a limited trial is available in the niji · journey app on iOS and Android (Midjourney). Plans start at $10 a month, and Stealth Mode, which keeps your images and videos private, is only available on the Pro and Mega plans (Midjourney).

Simple tips for writing a prompt

The description you write is called a prompt. These habits work with most tools:

  1. Start with the subject, then the place, then the style. For example: "a falcon on a falconer's glove, in the desert at dawn, in the style of an oil painting".
  2. Mention the light, the angle and the shape. "Soft light, close-up, portrait format for a phone screen."
  3. Change one thing at a time. If you don't like the result, ask for one change at a time instead of writing a new description.
  4. Try Arabic and English. You can write in Arabic. If you don't get what you want, try the same description in English and compare.

What about Arabic text inside the image? This is still one of the hardest parts. When OpenAI launched image generation in GPT-4o on 25 March 2025, it listed "Multilingual text rendering" among the model's known limitations (OpenAI). Google says its newer model adds text to images "with more accurate spelling, for all supported languages" (Google). In practice, check every letter before you share the image, because Arabic letters can come out disconnected, reversed or as words that do not exist. If the text matters, such as a shop name or an invitation card, ask for the image without text and add the text yourself in a design app that supports Arabic.

Before you upload your photo or your children's

A photo you upload is not just a file. It shows your face, and sometimes your home, your car or your child's school. These questions are worth a minute before you upload:

  • Is the photo used to improve the service? Google says that if you share a photo from a connected media gallery app with Gemini and your Keep Activity setting is on, the photo "is used to improve Google services with the help of human reviewers" (Google). OpenAI says it may use content from its individual plans to train its models, and that you can turn this off in settings with "Improve the model for everyone" (OpenAI).
  • Can other people see it? In some tools, creations are public by default. Leonardo.Ai's free plan describes creations as public (Leonardo.Ai), and Midjourney's private mode is only on its two top plans (Midjourney).
  • Does the background reveal more than you want? A number plate, a school name on a uniform, or the view from your window. Crop the photo before you upload it.

Children's photos need extra care. In the UAE, Federal Decree-Law No. 26 of 2025 on Child Digital Safety bars digital platforms from collecting, processing, publishing or sharing the personal data of children under 13, except under specific conditions. It also asks caregivers to monitor children's digital activity and use parental controls (Gulf News, 26 Dec 2025; Clyde & Co, Jan 2026). A simple rule for every family: do not upload a child's photo to a tool when you don't know what it does with images, and never share an edited photo of someone else's child. In October 2026 Dubai launched a free guide for parents on children's use of AI, which covers everyday situations including creating or sharing a fake image of a real person (our news item on GenDubAI).

For more habits that protect your data, read Is AI watching me? What happens to my data?.

Other people's images: what Gulf laws say

When you create an image of someone else or edit their photo, you are dealing with their right to privacy and to their reputation. These examples come from the UAE and Saudi Arabia. They are for general awareness and are not legal advice.

UAE. Article 44 of Federal Decree-Law No. 34 of 2021 on Countering Rumours and Cybercrimes punishes anyone who uses technology to invade a person's privacy without their consent. This includes taking, copying, keeping or publishing photos of other people. The penalty is imprisonment of at least six months and/or a fine of AED 150,000 to 500,000. The same article says that anyone who modifies or processes a photo with the intention of defaming or insulting another person faces imprisonment of at least one year and/or a fine of AED 250,000 to 500,000. The decree-law was amended in 2024, and the official text as updated on 4 April 2024 still sets out Article 44 in these terms (UAE Legislation; English translation of the decree-law, published by ADGM). In September 2026 the National Media Authority said in a post on X that using AI to generate images or videos of national symbols or public figures without prior official approval violates the UAE's media content standards, according to three UAE newspapers (Al Khaleej, 29 Sep 2026; Khaleej Times, 30 Sep 2026; Gulf News, 29 Sep 2026).

Saudi Arabia. Article 3 of the Anti-Cyber Crime Law, issued by Royal Decree No. M/17 of 8 Rabi I 1428H (26 March 2007), punishes invading privacy through the misuse of camera-equipped mobile phones and similar devices, and defaming or harming others through information technology. The penalty is imprisonment of up to one year and a fine of up to SAR 500,000, or either one (WIPO Lex). The Saudi Data and AI Authority (SDAIA) has also issued deepfakes guidelines. They ask creators to "obtain explicit consent from individuals whose likeness is used in the creation of deepfakes", to apply visible, tamper-resistant watermarks to deepfake content, and to tell audiences clearly that the technology was used (SDAIA, Deepfakes Guidelines). According to Arab News, which reported on the guidelines in May 2026, they also advise victims to document evidence immediately and report the content to the platform (Arab News, 9 May 2026).

Each of the other GCC states has its own cybercrime and data laws, which we do not cover here. A rule that works everywhere: do not create an image of a real person or edit their photo without their permission, and do not share an image that could embarrass them or mislead people about them. To learn how to spot a deepfake and avoid scams, see question 14 in our FAQ.

Labelling AI images

Labels help people know that an image is not a real photograph. There are several ways to do it:

  • Hidden data inside the file. OpenAI says all images generated by GPT-4o carry C2PA metadata that identifies them as coming from the model (OpenAI). Google said in April 2025 that images created or edited in Gemini carry its invisible SynthID watermark (Google, 30 Apr 2025), and its Gemini help pages now explain how to check an image for that watermark in the Gemini app (Google). But this data can be lost when someone takes a screenshot or re-saves the image.
  • Visible icons. In July 2025 the Dubai Future Foundation launched its Human–Machine Collaboration icons, five levels from "All Human" to "All Machine", and the Crown Prince of Dubai directed Dubai Government entities to use them in their research and knowledge work (Gulf News, 16 Jul 2025; Middle East AI News, 16 Jul 2025).
  • A word from you. The simplest label is to write "image made with AI" when you share it. SDAIA's guidelines ask creators for visible disclosures of this kind, such as a note in the description or text on the content itself (SDAIA).

What we don't know yet

  • The free limits in numbers. OpenAI does not publish how many images the free plan allows each day, and the limits change.
  • Where Leonardo.Ai and Midjourney work, country by country. We did not find a published country list from either company, so check when you sign up.
  • How well each tool handles Arabic text today. The models are improving fast, and we did not find a recent independent test comparing them on Arabic text inside images.

What this means for you

Creating images with AI can be an enjoyable hobby and a useful tool for a greeting card or an idea for a small business. Start with a tool you already have access to, write a clear description, and check any Arabic text before you share. Before you upload a face, ask where the photo will go and who will see it. Ask permission before you use anyone else's image, and say clearly that the picture was made with AI.

We checked each tool's facts on the company's own pages on 31 October 2026, unless stated otherwise.