Generate images from the sidebar

Say what is in the picture, where, and how it should feel. Specifics beat adjectives: a laptop on a wooden desk by a window, morning light, shallow depth of field works better than a beautiful workspace. Choose the shape for where it is going: Landscape for a blog header, Portrait for a story, Square for a profile picture.

The styles set the look so you do not have to spell it out. Logo and Social post are tuned for those uses; Photo aims for photographic realism. If the first result is close but not right, use Again and change one thing in the prompt at a time.

Image models are poor at long, accurate text inside pictures. If you need exact wording on a graphic, add it in a design tool afterwards.

Images you make are saved to Gallery, labelled All your images, so you can come back to one rather than regenerate it. On the web app at app.copilotly.com the same images appear in the AI Studio gallery. If you are working from a chat, the image sits in the chat history alongside the prompt that made it.

When you want a prompt written for you, ask the chat. For example, paste a product description and ask for three image prompts, then run them. If the picture should come from a document or a web page, use the page context button first so the chat knows what the page is about.

Images are counted per day, separately from your chat messages and credits:

• Free: 3 images a day. The chat shows 2 of 3 free images left today and when it resets.
• Pro: 100 HD images a day.

When you reach the Free limit you are offered Pro or the option to use your own key. Image limits are separate from the chat limits on your own API key, so do not assume a key lifts them. Pro costs $6.99/week, $19.99/month or $149.99/year, with a 7-day free trial for first-time subscribers. See pricing.

Before publishing generated images commercially, read the terms that apply to AI-generated content for the model and platform involved, and check whether your use needs disclosure. Avoid prompts that imitate a specific living artist, real people's likenesses or trademarked characters. This is general guidance, not legal advice.

Pair the picture with words from the rest of the product: write the caption in rewrite anywhere, or research the topic with web search first.

The jobs where a generated image earns its place are the ones where a stock photo would be generic and a commissioned illustration would be overkill:

• Blog and newsletter headers. Landscape shape, Illustration or Photo style, one clear subject and room at the edges for a title.
• Social posts. The Social post style, Square or Portrait, with a simple scene. Add any text yourself afterwards.
• Presentation visuals. A 3D or Illustration concept for a slide that would otherwise be a wall of bullets. You can also have Copilotly create the deck itself as a downloadable file.
• Concept sketches. Logo style for early ideas, to talk through with a designer rather than to ship.

What to skip: images that must show a real product, a real person or a real place accurately, charts and diagrams that need exact labels, and anything where a wrong detail would mislead. Image models invent details; a hand with six fingers is the famous example, but a wrong logo or an impossible building is just as common.

A good loop is three rounds at most. Generate, change one thing in the prompt, run Again, and pick the best. If the third round is not close, rewrite the prompt from scratch instead of piling on corrections. Because the free plan gives you 3 images a day, it pays to write a careful prompt first; if you want help, ask the chat for three variations of a prompt and run only the one you like.

Choosing shape and style first saves a regeneration, and on the free plan you only get 3 images a day.

Choosing a shape

Where it is going	Shape
Blog header, slide background, video thumbnail	Landscape
Story or phone wallpaper	Portrait
Profile picture, product tile, feed post	Square

Choosing a style

Photo aims for photographic realism, which suits scenes, objects and settings. Illustration and Watercolor give softer, editorial looks that hide small errors better than photos do. 3D suits product-style renders and slide concepts. Anime is a stylised look. Logo and Social post are tuned for those formats, and Auto lets the model decide when you are not sure.

When Auto is the right choice

If your prompt already describes the look in words, Auto avoids two instructions fighting each other. If you want to compare looks, pick a fixed style and change only that between runs.

Treat the first image as a draft of the prompt rather than a draft of the picture.

Change one thing at a time

If you change the subject, the lighting and the style together you will not know which change helped. Alter one element, run Again, and compare.

Say what to leave out

Generic prompts attract generic extras. Be specific about the setting, and say what should not be there, such as no people, no text or a plain background.

Know when to start over

If three rounds have not got closer, rewrite the prompt from scratch. Piling corrections onto a prompt that has gone wrong usually makes it worse. Ask the chat to write three alternative prompts from your description, run the best one, and keep the rest for later.

Add alt text when you publish

A generated image on a page still needs a description for people using screen readers. The W3C guide to image accessibility explains how to write it, and the chat can draft a first version if you describe the picture.

Often yes, but three questions decide it, and none of them is answered by Copilotly alone.

Who owns it?

Copyright for AI-generated material is unsettled and differs by country. The US Copyright Office publishes its position on AI and copyrightability, and it is worth reading before you build a business on an image you cannot protect.

Does the content rule apply?

Check the terms of the model provider and any platform where you publish, and whether disclosure is expected. Search engines accept AI-generated content in general, but Google's guidance warns against producing content at scale mainly to manipulate rankings.

Does the image show something real?

Avoid real people's likenesses, living artists' styles and trademarked characters. A logo should be checked against existing marks before it goes on a shop front. This is general guidance, not legal advice.

Most wasted images come from skipping the planning step. A short routine keeps the three free images a day for pictures you will actually use.

Decide the job first

Write one sentence: what the image is for, where it will appear and what size it needs to be. That answers the shape question and tells you how much detail matters. A thumbnail does not need the fine detail that a full-width header does.

Draft the prompt in the chat

Describe the page or the post and ask the chat for three image prompts in different styles. Pick one, adjust the details and run it. This is also where the web search feature helps if you need to know what something really looks like before you describe it, such as a historical building or a type of plant.

Generate, review and keep

Look at the whole image, then the edges and the details. Check hands, text, reflections and anything that is meant to be accurate. Download the winner, which saves it to your device, and leave the rest in Gallery.

Edit outside Copilotly

Crop, add text, adjust colours and compress in a design or photo tool. That is the stage where a generated picture becomes a finished asset, and it is where exact wording, logos and brand colours should be added.

Write the words around it

Pair the picture with copy you have checked. For example, select a draft caption on the page and use rewrite anywhere to adjust tone or length. If the image supports an article that needs sources, run deep research first so the picture and the text agree.

Dedicated image tools offer more control than a sidebar can. Be honest about which one you need.

What a dedicated tool does better

Specialist image apps typically offer finer controls such as editing part of an image, consistent characters across a series, upscaling, and large batches. If your work is design, illustration or marketing at volume, use one of those.

What the built-in generator does well

It is already where you are working. You can read a page, ask for an image that fits it, download it and carry on without switching apps or setting up another account. There are eight styles, including ones for logos and social posts, and a gallery that keeps the results.

What to check before relying on it

Question	Answer in Copilotly
Can I choose a shape?	Yes: Square, Landscape, Portrait
Can I choose a style?	Yes: eight, including Logo and Social post
Can I edit part of an image?	Not in the steps this page describes. Use an image editor
Can I keep what I made?	Yes: Download, Copy image and Gallery
Does my own API key lift the limit?	No

For the models behind the chat, see the model pages, and for the plan details see pricing.