AI Image Generation Glossary: Key Terms Explained

AI image generation has its own vocabulary, and most of it sounds more technical than it is. This glossary defines the terms you will actually encounter — what they mean, and why they matter when you are trying to get a specific image out of a model. Use it as a reference; you do not need to memorize it to start creating images.

Core concepts

These are the foundations — the words that come up in almost every prompt or settings panel.

• Prompt — The text description you give the model. It names the subject, style, lighting, and composition. The prompt is your main steering wheel.
• Negative prompt — A description of what you do not want in the image (for example, "blurry, extra fingers, text"). Some tools use it to steer the model away from common errors.
• Seed — A number that fixes the random starting point of a generation. Same prompt plus same seed equals a near-identical image, which is how you reproduce or fine-tune a look.
• Diffusion — The underlying method behind most modern image models. The model begins with random noise and refines it step by step until a coherent image emerges that matches your prompt.
• Inference / generation — A single run of the model that produces one or more images from your prompt.

Image quality and output terms

These describe the image itself — how big it is, how it is shaped, and how detailed.

• Resolution — The pixel dimensions of the image, like 1024 x 1024 or 4096 x 2160. Higher resolution means more detail and larger usable files.
• Aspect ratio — The shape of the frame: 1:1 (square), 16:9 (widescreen), 9:16 (vertical), 4:5 (portrait feed). Choose it to match where the image will be used.
• Upscaling — Increasing an image's resolution while intelligently adding detail, so a small render becomes a sharp 4K or print-ready file.
• Detail / fidelity — How crisp and refined the image looks. Higher fidelity means cleaner edges, textures, and fine features.
• Artifact — A visual error introduced by the model — warped hands, garbled text, melted edges. Reprompting or a different seed usually fixes them.

Prompting and control terms

These give you finer command over what the model produces.

• Style pack / preset — A saved bundle of look-and-feel instructions (such as cinematic, anime, watercolor) you apply with one click instead of writing the style out each time. eaxy ships 30+ of these.
• Reference image — An image you provide to guide the output's composition, subject, or style, rather than starting from text alone.
• Image-to-image — Generating a new image using an existing image plus a prompt as the starting point, so the result keeps some structure from the original.
• Weighting — Emphasizing parts of a prompt so the model prioritizes them (for example, making "golden light" stronger than "indoor").
• Consistency — Keeping a character, object, or style the same across multiple images — important for series, comics, and brand work.

Models and motion terms

The wider ecosystem you will hear named when comparing tools.

• Foundation model — The large trained model doing the generation. Well-known image models in 2026 include Midjourney v7, DALL-E 3, Adobe Firefly, Leonardo, and Ideogram (strong at rendering text).
• Text-to-image — Generating an image from a written prompt alone. See text-to-image explained for the full breakdown.
• Image-to-video — Animating a still image into a short clip by adding motion. The leading 2026 video models include Kling 3 (used by eaxy), Runway Gen-4.5, and Google Veo 3.1.
• Text-to-video — Generating a moving clip directly from a written prompt, no source image required.
• Commercial license — The right to use generated images and clips for business purposes. eaxy includes this on Pro and above.

How to use this glossary

You do not need to learn all of this before you start. The fastest way to make these terms click is to generate a few images and watch what each setting does:

1. Write a prompt and generate — that is a prompt and an inference.
2. Change the aspect ratio and regenerate to see the frame reshape.
3. Reuse the seed with a small prompt edit to see controlled variation.
4. Apply a style pack to change the entire look in one click.
5. Upscale your favorite to a high-resolution export.

Once the words map to actions, prompting stops feeling like guesswork. For the next step, read the AI image prompting guide, then start creating and put the vocabulary to work.