Your Everyday Glossary for Planning AI-Assisted Video

If you're just getting started with AI video tools, the vocabulary can feel like a lot. Some of it comes from old-school filmmaking, and some of it is brand new. Here's a plain-English walkthrough of the terms you'll run into, in roughly the order you'll need them, plus a simple routine for checking your work before anyone else sees it.

Planning from idea to clip

Think of planning as zooming in. You start with the story, then the scenes, then the shots, then the tiny frames inside each shot.

Scene. A scene is one moment in one place, like a woman sipping coffee by a window. Before you open any tool, jot down your whole video as a short list of scenes. Magic Hour's FAQ says its text and image modes make individual clips, not finished videos with multiple scenes, narration and captions. So stitching scenes together is on you.

Shot. A shot is one continuous take from one camera position or movement. Here's an easy test. If your description needs the word "then" more than twice, you've probably got two shots hiding in one. Shorter, simpler shots are easier to compare and fix.

Prompt. A prompt is the written instruction you give the tool. It covers your subject, the movement, the camera and the mood. Be specific, but don't expect a perfect match every time. Treat each result as a rough draft.

Aspect ratio and resolution. The aspect ratio is the shape of your picture. 16:9 is wide, 9:16 is tall for phone feeds, and 1:1 is square. Resolution, like 480p or 1080p, is how sharp it looks. Pick both before you generate, because changing the shape later can chop off something important. Magic Hour notes that the options you get depend on the model and your plan.

Frame and frame rate. A frame is a single still image. Video is just a lot of them shown fast. Frame rate, measured in frames per second, affects how smooth the motion feels. When someone tells you to check a clip "frame by frame," they mean pausing on individual stills to catch flaws your eyes skip at full speed.

Start and end frames. These are the first and last images of a clip. Some models let you lock them in, which helps when your clip has to open or close on a specific picture. Whether you can do that depends on the model you pick.

The AI-specific terms

Image to video. This means giving the tool a still picture as the starting point and letting it generate movement from there. Magic Hour describes a simple flow: upload an image, optionally describe the motion you want, then generate and preview. If you're learning, an image to video AI workflow is a friendly place to begin because you control how the clip opens. Just keep an eye on faces, hands, text and small objects, since those can drift once things start moving.

Motion prompt. This is the part of your prompt that describes movement, like "camera slowly pans right" or "hair moves in a light breeze." Stick to one or two actions per clip. Cramming in more makes odd results more likely.

Lip sync. Lip sync matches a speaker's mouth movements to a separate audio track. People use it to change a spoken line without reshooting, or to pair footage with a different language track. The publisher says results depend on how visible the face is, the quality of your source and how clear your audio is. It also points out that lip sync adjusts the video to fit your audio but does not translate anything. For a closer look at how it works, check out lip sync AI in Magic Hour, especially the tips about using front-facing footage with one clearly visible speaker.

Native audio. This is sound generated along with the picture, as opposed to audio you add afterward. Support changes from model to model, so confirm what your pick offers before you build a plan around it.

Video generation platform. This just means one workspace that handles several kinds of creation, like text to video, image to video, editing and lip sync, so you aren't bouncing between separate sites. Browsing an AI video generator from Magic Hour shows how that can look, including a table that compares models. The page labels that table as editorial guidance and not benchmark scores, so read it that way too.

Watermark. A watermark is a visible mark on your output. Free tiers often include one, and paid tiers may remove it. Losing the watermark does not erase your responsibility to tell viewers about AI use.

The review terms

Draft and review. Generated clips are first drafts. Plan on at least one round of regenerating.

Artifact. An artifact is a visual glitch caused by generation or compression, like warped fingers, flickering textures, melting edges or garbled lettering. Magic Hour's own pages warn that generated on-screen text may be inaccurate.

Continuity. This is whether people, outfits, props and settings stay consistent from frame to frame and from clip to clip. Problems usually show up when you join several generated clips together.

Disclosure. Disclosure means telling viewers when content is AI-generated or AI-altered. Rules vary by platform, client and location. Magic Hour's FAQ itself says that a missing watermark doesn't make a video undetectable or cancel disclosure duties, and that you should follow the labeling rules wherever you publish.

Rights. Rights are your permission to use the inputs and the outputs. Only upload images, footage and audio you're allowed to use, and read the tool's terms on commercial use before putting a clip in an ad or client project.

A simple review routine

1. Check your source first. Look over your starting image or footage before you generate anything. Dark lighting, heavy compression or a hidden face can all hurt your results. The publisher recommends high-resolution images with a clearly visible subject.
2. Watch it once at full speed. Notice your gut reaction. Does anything feel off?
3. Scrub frame by frame. Pause on faces, hands, edges and any text. Look for flicker, warping and sudden changes.
4. Check mouth against audio. For lip-synced clips, watch with sound, then with the sound off, then listen without looking. Timing drift is easier to catch when you split the senses.
5. Listen on its own. Make sure the speech is clear, pronounced correctly and says exactly what you meant. If it's a translated line, have a fluent speaker confirm the wording.
6. Test the format. Play it on a phone if your target is vertical video. Cropping can hide or cut off key details.
7. Verify claims. If your clip shows a product or states something as fact, make sure it's accurate and not misleading.
8. Confirm rights and labels. Recheck permissions for every asset, then add whatever AI disclosure your platform or client requires.
9. Keep notes. Write down the prompt, model and settings behind the version you approved so you can repeat or tweak it later.

One last thing

Most AI video mistakes come from skipping the plan or skipping the review. If you can describe your scenes, shots and frames in plain words, and you review each result the way an editor would, you'll catch most problems long before your audience does.