The Complete Guide to Creating Images with AI: Gemini, Bing and the Best Prompts

With the help of AI, you can now create an image just by typing a few words: a YouTube thumbnail background, a poster for your shop, characters for a story, or a social media post. In this article you will learn which free tools are available, how to write a good prompt, how to edit the image you get, how to keep the same character looking consistent again and again, and what you need to remember about usage rights and responsibility.

Note: The limits of free tools (how many images per day) and their features change often, so we are not giving any fixed numbers here. Check the current status of each tool on its own Help page.

Free AI image tools

Google Gemini

In the Gemini app or at gemini.google.com you can simply write "create an image of…". Google's image model (commonly known as Nano Banana) not only creates new images but can also edit an image you upload through conversation, for example "make the background blue" or "put a book in this person's hand". Free users have a daily limit, which changes from time to time. Google adds an invisible SynthID watermark to images made by Gemini, which indicates that the image was made with AI.

Microsoft Designer and Bing Image Creator

Both of these Microsoft tools can be used for free with a Microsoft account. In Bing Image Creator (bing.com/create) you type a prompt and images are generated. In Microsoft Designer, along with images, you can also create designs such as posters, social media posts and invitation cards. Microsoft is steadily integrating these tools with Copilot, so their interface and names may change.

Other options

• ChatGPT can also create images, with a limited allowance for free users.
• Canva has an AI image generator, and you can do the design work right there too.
• Adobe Firefly gives you some credits on its free plan.

The structure of a good prompt

Most people type "a beautiful village" and end up disappointed. For a good image, include these five parts in your prompt:

1. Subject: What or who? Age, clothing, action. For example, "an elderly farmer harvesting a wheat crop".
2. Style: A realistic photo (photorealistic), cartoon, 3D, watercolor, flat vector and so on.
3. Lighting: Soft morning sunlight, golden evening light (golden hour), studio lighting, neon.
4. Composition: Close-up, a wide shot of the whole scene, from above (top view), subject on the left with empty space on the right.
5. Aspect ratio: 16:9 for YouTube, 9:16 for Shorts and stories, 1:1 for an Instagram post.

For example:

A photorealistic image of an elderly Pakistani farmer harvesting golden wheat, warm golden hour light, wide shot with the farmer on the left and empty sky on the right for text, 16:9 aspect ratio, no text in the image.

In this prompt we specified all five: subject, style, lighting, composition and aspect ratio, and we also asked for no text in the image. AI models often write Urdu or English text inside images incorrectly, so it is better to add the text yourself later in a tool like Canva. You will find more ways to improve your prompts in our article Prompt Engineering in Urdu.

Another example: a shop poster

A clean flat vector illustration of a small Pakistani bakery counter with fresh naan and cakes, bright orange and white colors, friendly cartoon style, square 1:1, plenty of empty space at the top.

Editing the image

The first image is rarely perfect. To improve it:

• Edit through conversation: In tools like Gemini, write in the same chat: "keep everything the same, just add clouds to the sky". Ask for one change at a time.
• Change one word and try again: If you don't like the style, change only the style part and keep the rest of the prompt the same.
• Negative instructions: Phrases like "no text", "no watermark" and "no extra fingers" help in some tools.
• Final touch: Crop, adjust colors and add text in Canva or Photopea (a free editor that runs in your browser).

Keeping characters consistent (Character consistency)

When making a story or a series, the biggest problem is that the character's face changes in every image. To deal with this:

1. First create a character sheet: the character's name, age, facial features, hair, clothing colors and distinctive marks. Repeat this description exactly the same way in every prompt.
2. Once you get a good image, upload it as a reference image and say "now show this same character in a market". Tools like Gemini keep the character more consistent this way.
3. Keep the clothing and style the same in every image.
4. Even so, don't expect one hundred percent consistency every time; generate a few images and pick the best one.

Usage rights

• Every tool's Terms of Service states whether you can use the images you create for commercial work. Read these terms before any business use.
• In many countries, copyright over an image made entirely by AI is not clear, which means someone else might also use a similar image.
• Making an exact copy of a famous cartoon character, a brand logo or a particular artist's style can cause problems.
• Copyright rules on YouTube are explained in detail in our article YouTube Copyright Guide.

Images of real people: responsibility

Using AI to create a fake image of a real person and presenting it as if the event actually happened is deception, and in many countries it can also be a legal offence.

• Don't create an image of any person (especially a politician, an actor or an ordinary individual) that damages their reputation or misleads people.
• Editing someone's photo and spreading it without their permission is wrong.
• Use your own photo or the photo of someone who has given permission, and where needed, state that the image was created or altered with AI. Platforms like YouTube ask creators to disclose altered content that looks realistic.
• To protect yourself from fraud carried out with fake images, be sure to read Safety and Privacy in the Age of AI.

If you are creating images for YouTube, our guide Thumbnail Design with AI is your next step.

Summary

• Gemini, Microsoft Designer, Bing Image Creator, Canva and other tools are enough to get started for free, but their limits keep changing.
• Include the subject, style, lighting, composition and aspect ratio in a good prompt.
• Instead of having AI write text in the image, add it yourself afterwards.
• Use a character sheet and a reference image to keep a character consistent.
• Read the terms of use, and never create a misleading image of a real person.