Glossary
AI video glossary
Plain definitions of the terms behind AI video, AI characters and augmented reality.
Short, plain definitions of the key terms in AI video and augmented reality, as we use them at HAJFLIX.
Terms
Generative video
Video created by an AI model from inputs such as text, an image or another video, instead of being filmed with a camera.
Text-to-video
Generating video from a written description alone. Fast for testing ideas, but control over look and movement is limited.
Image-to-video
Animating a still image into a video shot. We use it, for example, to turn architectural renders into camera moves.
Reference image
One still image that defines a character's look: face, age, outfit and details. The model returns to it in every shot to keep the character consistent.
Character consistency
Keeping a character's face, age and details the same across every video and shot. It is the hardest part of AI character video.
Driving video
A recording of a real person performing the scene on camera. Their performance, timing, gestures and mouth shapes, is transferred onto the generated character.
Motion Control
A feature in tools such as Kling AI that transfers movement from a real driving video onto a generated character while keeping its look from the reference image.
Lip-sync
Matching a character's mouth movement to the audio. When it is off, viewers notice immediately.
Prompt
The written description that guides an AI model. It matters, but it can't make up for good inputs such as a strong reference image and a real performance.
AI avatar
A digital character that reads a script with synced voice and mouth movement, often used for presenter and explainer videos.
3D model
A digital object of a character or product that can be seen from every angle, used in augmented reality and 3D graphics.
Image-to-3D
Turning a still image into a 3D model with AI, on platforms such as Meshy, Tripo and Rodin.
AI augmented reality
Placing a 3D character or object into a real place through a phone camera, then generating a video in which it takes the place's light and shadows, so it looks filmed there. Unlike a filter, which only draws a layer over the camera. A real example.
Regeneration (video-to-video)
Using an existing shot as a base and generating it again with another video model, such as Seedance, Google Veo or Kling AI, for more control over movement, light or detail.
Vertical video (9:16)
The upright video format used on TikTok, Instagram Reels and YouTube Shorts.
Code-based motion graphics
Titles and animation written in code, with tools such as Remotion and HyperFrames, so text, colours or numbers can change and the video re-renders instead of being redesigned from scratch.
Have a project in mind?
Talk to the Haj