AI image glossary

This glossary explains the concepts that recur in Muse Image and in most AI image tools. Each entry is one paragraph and assumes no design or machine-learning background. Entries are ordered by how often they come up, not alphabetically.

Why the vocabulary is worth pinning down

Most frustration with AI image tools comes from mismatched terms: the "style" you mean is not the "style" the tool acts on, and people assume a negative prompt can fix composition when it can only exclude. Aligning on these words makes describing what you want measurably faster.

Aspect ratio: The proportion of width to height, such as 16:9, 3:4, or 1:1. It determines composition — where the subject sits, how much margin exists, whether the image survives as a thumbnail. Ratio and pixel size are different things: a high-resolution image at the right ratio crops to any target size, while a wrong ratio always costs you part of the frame.
Prompt: The text description handed to the image model describing what should appear. In Muse Image you do not write it directly — you fill in structured fields such as topic, audience, and purpose, and the system organises them into a prompt across ten visual elements.
Negative prompt: Text telling the model what must not appear, such as watermarks, stray lettering, or distorted hands. It can only exclude; it cannot fix composition or add to the positive description. Muse Image writes one automatically and merges in the avoid terms you supply.
Muse profile: Muse Image's style memory. It stores style tags, color hints, hard constraints, and a style summary, and appends them deterministically to the end of each prompt — positioned after your per-run intent, so it supplements rather than overrides what you asked for.
Card draw: Iterating from an image that has already been generated. Muse Image reads the actual image through a vision-capable model, assesses it across style, color, composition, mood, and typography, rewrites the prompt accordingly, and links the new image into the same lineage.
Credit: The unit of generation allowance. One generation costs 1 credit and produces 1 image, because each job calls the image model exactly once. Free accounts get 5 credits per month and Pro gets 60. Failed jobs refund the credit automatically.
OG image (Open Graph image): The preview image shown when a page is shared on social platforms, specified by the og:image tag in the page's HTML. The de facto standard size is 1200×630 (1.91:1). Note that the Open Graph protocol itself specifies no dimensions — that number comes from platform recommendations.
Safe area: The central region of a frame that survives the various crops a platform may apply. Keeping the subject and key elements inside it is what lets one image work across placements. Anything pressed against an edge is the first thing lost on mobile or in a square crop.
Hero image: The main visual at the top of a landing page, visible without scrolling. Usually a wide landscape image that needs clean negative space for an overlaid headline and call to action. Muse Image's Product Studio outputs a fixed 16:9 hero and reserves that space by default.
Variant: A different version of the same visual direction, used to choose among close options or to fill different placements. Variants share a style language but differ in composition or detail — which is not the same as generating an entirely different image.
Canvas: The pixel dimensions the model actually renders. Muse Image has three: 1536×1024 landscape, 1024×1536 portrait, and 1024×1024 square. Your chosen aspect ratio drives composition and lands on the closest canvas; crop to the exact target size on delivery.
Prompt moderation: The compliance check run against a prompt before the image model is called. When moderation denies, flags, or is unavailable, the request is blocked and never reaches image generation. Muse Image runs this on the final prompt of every generation.
Idempotency key: A unique identifier for a request that guarantees a retry will not produce a duplicate result. Muse Image reuses the same key when recovering from a gateway timeout, so a request the provider already accepted can be collected without charging twice.
Visual metaphor: Using one concrete object or material to carry an abstract product value — an opened darkroom door standing in for the start of creative work, for example. For software with no physical form, a visual metaphor is often more effective than depicting an interface.
Crop-safe: A composition requirement: the subject and its breathing room sit inside the centre of the frame, so the image stays intact across 16:9, square, and link-preview crops and remains legible at thumbnail size.
Style drift: The gradual loss of a shared visual language across images from one account or product line. It usually comes from describing the style slightly differently each time. Fixing the stable parts into a Muse profile addresses it directly.
Asset pack: The set of assets produced by one generation job, including the image and its role marker such as primary cover or hero. Packs carry purpose information into the design and publishing steps that follow, instead of leaving you with meaninglessly named downloads.
Lineage: The parent-child chain linking images produced through successive card draws. Each new image records the one it came from, so you can retrace the whole refinement path rather than being left with only the final version.

Can I use Muse Image without knowing these terms?

Yes. Muse Image is built on the premise that you should not need to write prompts — filling in the topic, audience, and purpose is enough. This glossary is for when you want finer control and need to know which part of the process to adjust.

What is the difference between a prompt and a negative prompt?

A prompt describes what should be in the image; a negative prompt describes what should not. A negative prompt can only exclude — it cannot fix composition or add to the positive description. To change composition, edit the positive description or use card draw.

How do aspect ratio and pixel size differ?

Aspect ratio is a proportion and drives composition; pixel size is an absolute measurement and drives sharpness. Getting the ratio right matters most at generation time, because a high-resolution image at the correct ratio crops to any size, whereas a wrong ratio costs you image content.