AI image & video glossary

Plain-language definitions of the terms used across AI image and video generation.

AI image generation

The creation of new images by an artificial-intelligence model from a text description, a reference image, or both, rather than by a camera or a human illustrator.

Generative AI

A class of AI models that produce new content — images, video, text — rather than only classifying or analysing existing data. Image and video generation are generative AI.

Text-to-image

Generating an image purely from a written description (a prompt), with no reference photo. Useful when you start from an idea rather than an existing object or space.

Image-to-image

Generating a new image guided by an existing one. The reference image constrains the result — its subject, composition, or proportions — while the model changes style, background, or other elements.

Image-to-video

Generating a short video clip from one or more still images. The model composes motion, animating a single frame or interpolating between several.

Reference image

A photo supplied to the model as a starting point or constraint — a room to restyle, a product to re-photograph, a frame to animate. It keeps the result faithful to a real subject.

Prompt

The text instruction given to an AI model describing what to generate. A clear, specific prompt produces a more predictable result.

Photorealistic render

A generated image intended to look like a real photograph rather than an illustration. yalmai’s interior and architecture flows aim for photorealistic output.

Virtual staging

Showing an empty or dated property furnished and styled, using generated images instead of physically furnishing the space. Common in real-estate marketing.

Credit

yalmai’s unit of usage. Each generation costs a number of credits depending on type, size, and quality. Credits are bought in packs and do not expire.

Flow

On yalmai, a flow is a generation mode tuned for a specific use case — interior design, product photography, architecture, or video — so you do not have to engineer a prompt from scratch.