How to Make AI Images: Every Chatbot Compared

Make the same 3 images (a birthday card, a product photo, and an edit of your own picture) in each chatbot, and know which one to use for what.

Checked October 17, 2026. Researched and drafted with help from Grok Bot.

Quick answer

• Five chatbots make images for free (with limits): ChatGPT, Gemini, Grok, Meta AI and Microsoft Copilot.
• Just ask in plain words. Type "Create an image of…" and describe it. Most apps also have an image button.
• You can edit your own photos too. Upload a picture and say what to change. ChatGPT, Gemini, Grok, Meta AI and Copilot all support this.
• Midjourney is the art-quality step up. It's made only for images (and short videos), but there's no free plan.
• Claude and DeepSeek don't make images. Perplexity does, but only on its paid Pro plan (not verified this week).

What's a "watermark"? A mark, visible or hidden, that shows a picture was made by AI. Several apps add one automatically.

Comparison table

Chatbot	Image tool name	Free?	Edit your own photo	Words inside images	Watermark	Best for
ChatGPT	ChatGPT Images (version 2.5)	✓ limited; "Images with thinking" on Plus, Pro, Business	✓ with a Select tool for one area	✓ follows instructions to add text	Not checked	Posters, cards, edits with clear instructions
Gemini	Nano Banana (free) / Nano Banana 2.1 (paid)	✓ limited	✓ including combining 2 photos	✓ Google highlights clear text	Hidden SynthID + visible mark	Photo edits, restyling, resizing
Grok	Grok Imagine	✓ limited; video on Premium	✓ restyle and edit from reference photos	Not documented	Visible Grok watermark (can't be removed)	Fun, bold images, and turning them into video
Meta AI	Muse Image	✓ free for everyday use	✓ upload and describe a change	Not documented	Hidden Content Seal	Quick images inside apps you already use
Microsoft Copilot	Create an image (in Create)	✓ subject to capacity	✓ add a reference image; Edit tools	Has a Text tool	Not checked	Style presets and set sizes
Midjourney	V8.2	✗ from $10/month	✓ Editor	✓	Public gallery unless Stealth Mode (Pro, Mega)	Art-quality, consistent style
Perplexity	Image generation	✗ Pro only	Not checked	Not checked	Not checked	If you already pay for Pro
Claude, DeepSeek	None	✗	✗	✗	—	Not for images

Midjourney price: Basic $10/month (US, checked Oct 2026, docs.midjourney.com).

Which one should I use? The verdict

Quick questions

Can I make AI images for free? Yes. ChatGPT, Gemini, Grok, Meta AI and Copilot all include image making on their free plans, with limits that reset over time.

Which AI is best for text in images, like a birthday card? ChatGPT and Gemini are the safest picks. Both companies say their tools follow instructions for adding text. Always check the spelling before you print.

Can ChatGPT still make videos? No. OpenAI shut down the Sora app and website on April 26, 2026, and the Sora API on September 24, 2026. For video, see our upcoming Advanced guide on AI video.

Can I make an AI picture of myself? Yes, in several apps. Upload your photo and describe the change. In Meta AI, you can include "me" or "myself" in your prompt. Only edit other people's photos with their permission.

Can I use AI images for my business? Often yes, but rules differ by app. Midjourney, for example, requires companies earning over $1 million a year to be on its Pro or Mega plan. Read each app's terms before you sell anything.

Before you start

• Sign in. Most apps need an account to make images. Gemini, for example, doesn't offer image creation when you're signed out.
• Check limits. Free plans limit how many images you can make. If you hit a limit, wait for it to reset or try another app.
• Privacy reminder: uploading a photo means sharing it with the company. Don't upload photos of documents, ID cards, or other people's kids. See our free guide AI Privacy and Fact-Checking (coming soon).

The 3 test images

Use these prompts in every app so you can compare fairly.

Test 1: a birthday card (tests text)

Create a square birthday card: a cartoon cat in a party hat holding balloons, pastel colors, flat illustration style. Write "Happy Birthday, Mia!" across the top in a fun, bold font. Check the spelling.

Test 2: a product photo (tests realism)

Create a wide product photo of a handmade lavender soap bar on a white marble counter, soft morning light from a window, a sprig of lavender beside it, shallow depth of field, clean and bright.

Test 3: edit your own photo (tests editing)

Upload a photo of a room, pet or object (not a person's face for this test), then:

Keep everything the same, but change the background to a sunny beach and make the lighting warm golden hour.

What good looks like: Test 1 has correctly spelled text. Test 2 looks like a real photo. Test 3 keeps your subject the same and only changes what you asked.

Step by step: ChatGPT

1. Go to chatgpt.com or open the app. Sign in.
2. Either type "Create an image of…" in the chat, or select More and then Images.
3. Describe your image. Include subject, style, colors and shape.
4. Wait. OpenAI says it can take a few minutes for complex images, and you can keep chatting meanwhile.
5. To change the shape: use the aspect ratio picker, or say "make it wide."
6. To edit one area: select the image, choose Select, highlight the part to change, and describe the change. OpenAI notes highlights aren't always precise.
7. To save: open the image and choose Save (download), Copy or Share.

Extras:

• Templates: open the sidebar, select Images, then Templates for starting points like posters or logos.
• Sketch (mobile app): type @, choose Sketch, draw, then add instructions like "turn this into a watercolor painting."
• Transparent background: just ask for one.
• Find old images: everything you make is saved under Images.

Plans: ChatGPT Images is on all plans. Images with thinking (more careful planning before drawing) is on Plus, Pro and Business.

Step by step: Gemini

1. Go to gemini.google.com or open the Gemini app. Sign in.
2. Select 🍌 Create images from the tools menu.
3. Type your prompt, or upload an image to edit.
4. Ask for changes in plain words: "change the background," "replace the cup with a mug," "make it night."
5. Combine photos: upload two pictures and ask Gemini to blend them into one.
6. Resize: ask for a different shape. Google says it can resize without cropping out the details.

Free vs paid: free users get Nano Banana. Google AI Plus, Pro and Ultra get Nano Banana 2.1, Google's best image model, with better text, character consistency and 2K resolution. On a paid plan, you can also open an image's three-dot menu and choose Redo with Pro to regenerate it with Nano Banana Pro.

Watermarks: Gemini images carry an invisible SynthID watermark and a visible one. You can upload any image to Gemini and ask if it was made with Google AI.

Good to know: Google's image page lists the feature as 18+, and says limits apply.

Step by step: Grok

Grok Imagine makes images and video from text or reference photos, and lets you restyle and edit without leaving the chat.

1. Open grok.com or the Grok app. Sign in.
2. Type "Create an image of…" and describe it.
3. To edit, upload a reference photo and say what to change: "make this a comic-book style," "change the season to winter."
4. Ask for one change at a time.
5. Download the version you like.

Plans: image generation is on the free plan with limits. Video generation starts at the Premium plan of The X Subscription ($30/month, US, checked Oct 2026, x.ai/pricing). SpaceXSI says Imagine goes up to 2K images and 15-second videos.

Watermark: Grok images and videos carry a Grok watermark. There's no setting to remove it, and removing it breaks SpaceXSI's rules.

If your video comes out blurry: 720p videos drop to 480p once you hit your plan's 720p cap.

Step by step: Meta AI

1. Open the Meta AI app (or meta.ai).
2. Tap Ask Meta AI… at the bottom.
3. Start your prompt with "imagine" or "create an image," and add details.
4. Tap Send. Meta AI makes several images.
5. Add yourself: include "me" or "myself" in the prompt. Add someone else: type "@" and their username (with their OK).
6. Make it move: tap an image, then Animate, to turn it into a short video Meta calls a "vibe."

Meta's own example shows how details help: "Imagine a polar bear at a pool" gives a plain picture. Adding "relaxing in a deck chair… other polar bears playing… early 2000s 3D animation style" gives a much better one.

Good to know: Meta uses your chats with its AI to improve its AI, and may use profile info to personalize. Meta's newer Muse Video model was only a preview in July 2026. Its current availability is not verified.

Step by step: Microsoft Copilot

1. Open Copilot in your browser or the Copilot app.
2. Select the Apps and more icon at the top left, then Create.
3. In What do you want to create?, choose Create an image.
4. Pick a template from Explore images, or type in Describe the image you want to create.
5. Optional: select +, then Add an image to use a reference picture.
6. Choose Style (like Photorealistic or Anime), Color, and Size: Square (1024 × 1024), Portrait (1024 × 1536) or Wide (1536 × 1024).
7. Select the arrow to create.
8. Refine with Transform with AI, or the Edit and Text tools. Save with Save and copy. Find old images in Copilot Library.

Plans: free users can create images subject to capacity. Microsoft 365 Personal and Family include standard image usage. Premium and Pro include extensive usage.

Step by step: Midjourney

1. Go to midjourney.com and sign in. Choose a paid plan (there's no free plan).
2. Type your prompt in the Imagine bar.
3. Add the shape at the end, like --ar 3:2. V8.2 is the default version.
4. Press Enter. Midjourney makes a set of images.
5. Open one in the Editor to change part of it or extend it.
6. For a consistent look across many images, build a Personalization profile or a moodboard.

Good to know: images are public in the gallery unless you have Stealth Mode (Pro and Mega plans). Midjourney can also turn an image into a short video.

The others

Safety: real people, kids and copyright

• Real people: don't make fake images of real people that could fool or embarrass someone. Get permission before editing anyone else's photo.
• Kids: avoid uploading children's photos, especially other people's.
• Brands and characters: apps may refuse logos, famous characters or an artist's exact style. Don't try to get around it.
• Labels: if you share an AI image, say it's AI-made. Several apps add watermarks for this reason.

Common mistakes and fixes

Problem	Likely cause	Fix
Misspelled words in the image	AI text is still imperfect	Keep text short. Put the exact words in quotes. Ask "check the spelling." Try ChatGPT or Gemini
Weird hands or faces	Common AI weak spot	Ask for a fix to that area only, or use Select (ChatGPT) or the Editor (Midjourney)
Edit changed the whole picture	Prompt was too broad	Start with "Keep everything the same, except…"
Wrong shape	Shape not stated	Say "square," "wide" or "tall," or use the aspect ratio picker or Size option
"Limit reached"	Free-plan cap	Wait for the reset, or switch apps
Request refused	Safety rules (real people, brands, violence)	Rephrase to something fictional and general

Prompt pack

Create a [square/wide/tall] [type of image: poster, card, logo, photo] of [subject], [setting], in [style] style, with [colors]. Mood: [cozy/bold/calm].

Keep everything the same, but change only [one thing].

Make 3 different versions of this idea so I can pick one: [idea].

Create a simple logo for [business name], a [type of business]. Flat design, 2 colors: [color] and [color]. No tiny details. Write the name "[name]" clearly.

Turn this photo into a [watercolor/cartoon/vintage postcard] style. Keep the subject recognizable.

Create a flyer for [event] on [date] at [place]. Put this exact text on it: "[text]". Easy to read from far away.

Create a coloring page for kids: [subject], thick black outlines, white background, no shading.

Next level

• A2: How to Make AI Videos in 2026 turns one photo into a short clip in Gemini, Grok and Midjourney.
• I2: How to Talk to AI has the image-prompt formula: Subject · Setting · Style · Light and color · Shape.
• Beginner guides: ChatGPT, Gemini, Grok, Meta AI and Midjourney.