Terms you'll run into around AI image and video generation. These are general to the field, not specific to any one tool, and not every control described here is exposed in every tool.

Generation

Prompt: the text description you give the model. Negative prompt: a description of what you want kept out of the image, where a tool supports it.

Text-to-image: generating a picture from a written description alone. Image-to-image: using an existing picture as the starting point, so the result keeps some of its structure. Text-to-video and image-to-video are the same ideas for motion.

Seed: the number that determines the random starting point of a generation. Reusing a seed with the same prompt reproduces a near-identical result, which is what makes controlled comparison possible.

Steps: how many refinement passes the model makes. More steps generally means more detail and more time.

Batch: generating several variations from one prompt in a single run.

Models

Model: the trained system that produces the output. Different models have different strengths, speeds and costs.

Diffusion: the technique behind most current image models. It starts from visual noise and removes it step by step until an image matching the prompt emerges.

LoRA: a small add-on trained to push a base model toward a particular subject or style, without retraining the whole model.

Checkpoint: a saved version of a model's trained state. In practice, one selectable model.

Editing

Inpainting: regenerating a selected area inside an image while leaving the rest untouched. Used for removing objects and fixing details.

Outpainting: extending an image beyond its original edges, generating what would plausibly have been outside the frame.

Upscaling: increasing resolution while reconstructing detail, rather than just enlarging pixels.

Background removal: separating the subject from its background to produce a cut-out, usually on a transparent background.

Face swap: replacing the face in one image with a face from another.

Output

Aspect ratio: the shape of the frame, like 1:1 square, 16:9 widescreen or 9:16 vertical. Worth setting to match where the result will be posted.

Resolution: the pixel dimensions of the output. 4K refers to roughly 4,000 pixels across.

Artifact: a visual flaw introduced by generation: warped hands, garbled text, smeared edges.

Credits: the unit MagicShot bills generation in. Heavier jobs cost more. See plans and credits.