AI Video Glossary
Every AI video tool uses the same handful of words slightly differently, which makes comparing them harder than it should be. This glossary defines the terms as they are used across Avocadio, Runway, CapCut and InVideo, so a feature list reads as what it actually does. Each entry says what the term means and, where it matters, why it changes the result you get.
- Text to video
- Generating a clip from a written prompt alone, with no image input. You describe subject, action, camera and lighting; the model invents the footage. Best for mood shots, B-roll and concepts you have no footage for.
- Image to video
- Animating a still image — a product photo, a portrait, a landscape — into a short clip. Because the frame is already yours, the subject, colours and branding stay recognisable; the model only supplies motion. This is the reliable route for product and UGC-style ads.
- Prompt pack (seed pack)
- A set of ready-written prompts for one purpose, such as a product reveal, a seasonal campaign or a UGC-style hook. Packs remove the blank-box problem: you swap in your own subject instead of writing motion directions from scratch.
- Prompt adherence
- How closely the generated clip matches what you asked for. High adherence means the camera move, framing and mood you described actually appear; low adherence means the model drifted. Lower the number of simultaneous instructions to raise it.
- Camera motion
- The movement the model performs during the clip — orbit, push-in, pull-back, pan, tilt, handheld drift. In image-to-video, slow camera motion reads as cinematic, while fast subject motion is what deforms faces, labels and small details.
- Negative prompt
- Instructions describing what should not appear — warped text, extra fingers, flickering, morphing, distorted logos. Used to steer a model away from the artifacts it reaches for most often.
- Aspect ratio
- The shape of the frame. 9:16 (vertical) is the native format for TikTok, Reels and Shorts; 1:1 and 4:5 suit some feed placements; 16:9 stays the standard for YouTube and web embeds. Choose the placement first, then generate in that shape.
- UGC-style ad
- A short ad shot to feel like a customer filmed it — handheld, natural light, a person holding the product, casual voice. AI UGC ads recreate that look from a product photo instead of hiring a creator.
- Credits
- The unit most AI video apps charge per generation. Because a usable clip often needs several attempts, the number that matters is cost per usable clip, not cost per generation.
- Label fidelity
- Whether the product's name, logo and packaging text stay readable through the animation. It is the single biggest quality factor in product video, and it is mostly protected by slow camera moves and high-resolution source photos.
Put the terms to work: see how to create product videos with AI or the aspect ratio guide.