AI image and video glossary
Plain-language definitions for prompts, diffusion, image-to-video, upscaling, seeds, aspect ratios and more.
- Technique
Depth Map
A depth map is a grayscale image where each pixel's brightness represents its distance from the camera — bright pixels are close, dark pixels are far. AI models use depth maps to understand 3D structure in 2D images.
Read → - Model
Flux
Flux is a family of text-to-image AI models developed by Black Forest Labs, known for high prompt adherence, strong text rendering, and photorealistic output. It is a popular alternative to Stable Diffusion.
Read → - Video
Frame Rate
Frame rate is the number of individual images (frames) displayed per second in a video, measured in FPS. Higher frame rates produce smoother motion; lower frame rates create a more cinematic or stylized look.
Read → - Video
Interpolation
Interpolation is the process of generating intermediate frames between existing frames to create smoother motion or slow-motion effects in AI video. It synthesizes in-between frames by analyzing the movement between adjacent originals.
Read → - Video
Keyframe
A keyframe is a defining frame in an animation or video sequence that marks the start or end of a transition. AI video models use keyframes as anchor points to generate the motion between them.
Read → - Video
Motion Brush
Motion brush is a video editing tool that lets you paint areas of a frame to define which regions should move and which should stay still, giving precise control over AI video motion.
Read → - Model
Runway Gen-3
Runway Gen-3 Alpha is a text-to-video and image-to-video AI model by Runway, known for high-fidelity cinematic output, motion brush control, and strong temporal consistency.
Read → - Model
Sora
Sora is OpenAI's text-to-video AI model, capable of generating high-quality video clips up to a minute long from text prompts. It is known for cinematic quality and long-duration generation.
Read → - Video
Temporal Consistency
Temporal consistency is the degree to which an AI-generated video maintains visual coherence across frames — keeping the same subject, colors, and lighting from frame to frame without flickering or morphing.
Read → - Model
Veo
Veo is Google DeepMind's text-to-video AI model, capable of generating high-quality 1080p video clips from text prompts with strong cinematic coherence and physics understanding.
Read → - Video
Video-to-Video
Video-to-video is an AI technique that transforms an existing video — changing its style, content, or quality — while preserving the original motion, timing, and composition.
Read → - Definition
Aspect Ratio
Aspect ratio is the proportional relationship between an image's width and height, written as two numbers like 16:9, 1:1, or 9:16.
Read → - Definition
Batch Generation
Batch generation is creating multiple AI images at once from a prompt or set of prompts, producing several variations in a single run to compare and pick.
Read → - Definition
CFG Scale
CFG scale (classifier-free guidance) controls how strictly an AI image model follows your prompt versus generating freely — higher means more literal.
Read → - Definition
Checkpoint
A checkpoint is a saved snapshot of a fully trained AI image model's weights — the complete file the generator loads to turn prompts into images.
Read → - Definition
Commercial License
A commercial license is the legal permission to use AI-generated images and video for business and revenue-generating purposes.
Read → - Definition
ControlNet
ControlNet is an add-on for image models that guides generation with a structural input like an edge map, pose, or depth — for precise control.
Read → - Definition
Denoising
Denoising is the step-by-step process a diffusion model uses to turn random noise into a clear image guided by your prompt.
Read → - Definition
Diffusion Model
A diffusion model is a generative AI that creates images by starting from random noise and gradually denoising it into a coherent picture.
Read → - Definition
GAN
A GAN (Generative Adversarial Network) is an AI model where a generator and discriminator compete, training the generator to produce realistic synthetic data.
Read → - Definition
Generative AI
Generative AI is a class of models that create new content — images, video, text, audio — from learned patterns, in response to a prompt.
Read → - Definition
Image-to-Video
Image-to-video is AI that animates a still image into a short video clip, adding motion and camera movement while keeping the original picture.
Read → - Definition
Image-to-Image (img2img)
Image-to-image (img2img) is AI generation that uses an existing image plus a prompt as the starting point, transforming it while keeping chosen structure.
Read → - Definition
Inpainting
Inpainting is AI editing that regenerates a selected region of an image, replacing or repairing content while blending seamlessly with the rest.
Read → - Definition
Kling 3
Kling 3 is a generative AI video model that turns text prompts or still images into short, high-fidelity video clips with controllable motion.
Read → - Definition
Latent Space
Latent space is the compressed numerical map where AI models represent images as points, so similar concepts sit close together.
Read → - Definition
LoRA
LoRA (Low-Rank Adaptation) is a lightweight fine-tuning method that teaches an AI model a new style, character, or concept without retraining it fully.
Read → - Definition
Negative Prompt
A negative prompt tells an AI model what to leave out of an image, steering the result away from unwanted elements or artifacts.
Read → - Definition
Outpainting
Outpainting is AI editing that extends an image beyond its original edges, generating new surroundings that match the existing scene.
Read → - Definition
Photorealism
Photorealism is AI-generated imagery so lifelike it resembles a real photograph, with believable lighting, texture, depth, and detail.
Read → - Definition
Prompt
A prompt is the text instruction you give an AI model to describe the image or video you want it to generate.
Read → - Definition
Prompt Engineering
Prompt engineering is the practice of crafting and refining the text instructions given to an AI model to get more accurate, useful, on-target results.
Read → - Definition
Render
To render is to generate the final image or video from a model and inputs; a render is the produced output of that computation.
Read → - Definition
Seed
A seed is the starting number that determines the random noise an AI uses to generate an image, making results repeatable.
Read → - Definition
Stable Diffusion
Stable Diffusion is an open-source latent diffusion model that generates images from text by iteratively denoising in a compressed latent space.
Read → - Definition
Style Pack
A style pack is a curated, reusable set of visual settings — look, lighting and aesthetic — applied to AI generations for a consistent style.
Read → - Definition
Style Transfer
Style transfer is an AI technique that applies the visual look of one image — its colors, textures, brushwork — to the content of another.
Read → - Definition
Text-to-Image
Text-to-image is AI that turns a written prompt into an original picture, generating the image from noise rather than retrieving it.
Read → - Definition
Text-to-Video
Text-to-video is AI that turns a written prompt into a short moving video clip, generating motion, scenes, and timing from words alone.
Read → - Definition
Upscaling
Upscaling is increasing an image or video's resolution while adding plausible detail, using AI to enlarge it without looking soft or blocky.
Read → - Definition
Watermark
A watermark is a visible or embedded mark added to an image or video to indicate ownership, source or that it was AI-generated.
Read →
