The shape of the image: width compared to height, like 1:1 square or 16:9 wide.
What the AI does
Sets the frame shape, which changes how the scene is composed.
When to use it
Before generating, so the image fits where you will use it.
Where it works
Midjourney: --ar W:H at the end of the prompt. ChatGPT/DALL·E and Gemini: pick a size or describe the shape in words. AI Compass: use the Size chips; words in the prompt are only a hint.
Common mistakes
Writing '16:9' in the prompt and expecting every tool to obey it.
Picking a square image for a wide banner.
Example
A wide landscape of rolling hills at sunrise, wide horizontal composition
Portrait / Square / Landscape
Also written: portrait, square, landscape, vertical, horizontal
Whether the image is tall, square or wide.
What the AI does
Changes the frame shape and how the subject is placed.
When to use it
Pick tall for phone screens and stories, wide for banners and desktops.
Where it works
AI Compass: choose it with the Size chips on the image screen (which shapes show depends on the app's settings). Other tools have a size menu or understand it in words. Midjourney uses --ar.
Common mistakes
Confusing 'portrait orientation' with 'a portrait photo of a person'.
Expecting the shape to change after the image is made.
Example
A tall vertical image of a waterfall in a forest, portrait orientation
image size
Also written: resolution, size, pixels, 1024x1024
How many pixels wide and tall the image is.
What the AI does
Larger sizes hold more detail; most tools offer a small fixed list.
When to use it
When the image must be printed or shown large.
Where it works
Most generators offer a few fixed sizes in their settings. AI Compass: chosen with the Size chips, not by typing numbers in the prompt. Writing '4K' changes the style, not the real pixel size.
Common mistakes
Typing '8K' and expecting a bigger file.
Choosing a tiny size for print.
Example
A detailed city map illustration, clean and sharp
negative prompt
Also written: negative prompt, exclude, without, no text
A list of things you do NOT want in the image.
What the AI does
Tells the generator to avoid certain objects, styles or flaws.
When to use it
When unwanted things keep appearing, like text, extra people or blur.
Where it works
Stable Diffusion tools: a separate Negative prompt box. Midjourney: --no. ChatGPT/DALL·E, Gemini and AI Compass: no separate box; write it as a sentence, e.g. 'no text in the image'.
Common mistakes
Writing 'no cats' in tools that may read 'cats' and add them; phrase it as 'an empty room' instead.
Listing dozens of negatives at once.
Example
Write a Stable Diffusion prompt and a separate negative prompt for a clean product photo of a watch, avoiding text, watermarks and blur.
seed
Also written: seed, seed number, random seed
A number that sets the starting randomness of an image.
What the AI does
Using the same seed with the same prompt and settings gives a very similar image.
When to use it
When you want to change one word and compare results fairly.
Where it works
Midjourney: --seed followed by a number. Stable Diffusion tools: a Seed field. ChatGPT/DALL·E, Gemini and AI Compass: no seed control; each run is new.
Common mistakes
Expecting the same seed to match across different tools or model versions.
Changing many settings at once and blaming the seed.
Example
Explain how to use a seed in Stable Diffusion to compare two versions of the same portrait prompt, step by step.
reference image
Also written: reference image, image prompt, image-to-image, img2img
An image you give the AI as a guide for style, layout or subject.
What the AI does
Uses the picture as inspiration while following your text.
When to use it
When words are not enough to describe the look you want.
Where it works
Midjourney: an image link at the start of the prompt. Stable Diffusion: image-to-image. ChatGPT and Gemini: attach the image. AI Compass: upload a photo in chat and describe what to change.
Common mistakes
Uploading images you don't have the rights to use.
Expecting an exact copy of the reference.
Example
I am uploading a photo of my living room. Suggest how to describe a cozier version with warm lighting and plants.
upscale
Also written: upscale, upscaling, enlarge, super resolution
Making an image larger and sharper after it is made.
What the AI does
Adds pixels and detail so the image looks better at a bigger size.
When to use it
When you need to print or crop a generated image.
Where it works
Midjourney and many image tools have an Upscale button. AI Compass has no separate upscale step; pick the size before generating.
Common mistakes
Expecting upscaling to fix a bad composition.
Upscaling a tiny image many times.
Example
Explain the difference between upscaling and generating at a larger size, and when each makes sense.
variations
Also written: variations, vary, regenerate, re-roll
New versions of an image that keep the main idea but change details.
What the AI does
Produces similar images with small differences in pose, color or layout.
When to use it
When you like the idea but not the exact result.
Where it works
Midjourney: Vary buttons. Other tools: regenerate or ask for variations. AI Compass: generate again with the same prompt, or change one word at a time.
Common mistakes
Changing many words at once, so you can't tell what helped.
Expecting identical results from the same prompt.
Example
A cozy reading nook with a window seat and plants, soft afternoon light
inpainting
Also written: inpainting, generative fill, edit part of an image, vary region
Changing only one part of an image while keeping the rest.
What the AI does
Redraws the selected area to match your instruction and blend with the rest.
When to use it
To fix hands, remove objects or change one detail.
Where it works
Midjourney: Vary (Region). Adobe Firefly/Photoshop: Generative Fill. Stable Diffusion: inpainting with a mask. AI Compass: upload the photo in chat and describe the change.
Common mistakes
Describing the whole image instead of just the change.
Editing photos of people without their consent.
Example
I will upload a photo of my desk. Remove the cables and add a small green plant on the right, keeping everything else the same.
outpainting
Also written: outpainting, generative expand, zoom out, extend image
Extending an image beyond its original edges.
What the AI does
Invents new surroundings that continue the scene outward.
When to use it
To change the shape of an image or add more background.
Where it works
Midjourney: Zoom Out and Pan. Firefly/Photoshop: Generative Expand. Stable Diffusion: outpainting tools. AI Compass: no expand tool; describe the wider scene and generate again.
Common mistakes
Expecting the new areas to be real details of the place.
Extending too far in one step.
Example
Explain what outpainting is and give me a prompt that describes a wider version of a beach scene with more sky and sand.
image editing with AI
Also written: edit photo, photo editing, change my photo
Changing an existing photo by describing what you want.
What the AI does
Applies your instruction, like changing the background or lighting, to the picture.
When to use it
When you have a photo and want a quick change without manual editing.
Where it works
ChatGPT, Gemini, Firefly and many apps accept an image plus an instruction. AI Compass: upload a photo in chat and describe the change.
Common mistakes
Using other people's photos without permission.
Asking for many changes at once.
Example
I will upload a product photo. Replace the background with a clean white studio background and keep the product unchanged.
text in images
Also written: text in image, words in image, lettering, typography
Asking the AI to write words inside the picture.
What the AI does
Tries to draw the letters; short words work better than long sentences.
When to use it
For short titles, signs and labels.
Where it works
Plain words in any generator, including AI Compass. Put the exact text in quotes and keep it short. Spelling mistakes are still common.
Common mistakes
Asking for long paragraphs of text.
Not checking spelling before using the image.
Example
A wooden café sign that says "OPEN" in simple white letters, hanging on a brick wall
transparent background
Also written: transparent background, PNG, no background, cutout
An image where the background is see-through.
What the AI does
Some tools can make a cut-out; others draw a white or checkered background instead.
When to use it
For logos, stickers and product cut-outs.
Where it works
Some generators and editors have a transparent or remove-background option. AI Compass images always have a background; ask for a plain solid one and remove it in an editor.
Common mistakes
Writing 'transparent' and getting a fake checkered pattern.
Using a busy background that is hard to remove.
Example
A single red apple on a plain solid white background, centered, soft shadow
image quality
Also written: quality, HD, high quality, 4K, ultra detailed
How detailed and clean the generated image is.
What the AI does
Higher quality settings usually take longer and cost more; quality words change the style.
When to use it
When detail matters, such as for print or large screens.
Where it works
ChatGPT/DALL·E and other tools offer quality options in their settings; Midjourney uses --q. AI Compass sets image quality itself. Words like '4K' or 'ultra detailed' change the look, not the file size.
Common mistakes
Adding '8K, masterpiece, best quality' to every prompt.
Expecting quality words to fix an unclear prompt.
Example
A detailed close-up of a hummingbird drinking from a red flower, sharp focus
Enhance with AI
Also written: enhance prompt, improve prompt, magic prompt
An AI Compass button that rewrites a short idea into a detailed prompt.
What the AI does
Adds style, lighting, composition and detail to your idea before generating.
When to use it
When you have a short idea and want a richer result.
Where it works
AI Compass image screen. Other tools have similar 'improve prompt' or 'magic prompt' features.
Common mistakes
Not reading the rewritten prompt before generating.
Using it when you need exact control over every word.
Every guide and dictionary card is also in the AI Compass app, with saved items, search and “Try it” buttons that open chat, image or video with the example filled in. Reading is free; creating uses credits, and the price is shown before you start.