The words of AI video
The terms that come up in the course and on generation platforms, explained simply. No tool is named here: only what you need to understand.
- Prompt
- The text that describes what you ask a model for: subject, framing, lighting, style. It's the first thing you learn to write, and the one you never stop improving.
- Model
- The artificial intelligence engine that generates an image, a video, a voice or music. A single platform often offers several, each with its own strengths.
- Text to image
- Creating a still image from a prompt. In a production pipeline, it's often the step that sets the scenery and the characters.
- Image to video
- Animating a still image: the model starts from the image you provide and gives it motion. Often safer than text alone for keeping a character identical.
- Reference image
- An image given to the model so it reuses a face, a style or a setting. It's the main lever for consistency.
- Consistency
- Keeping a character, a setting or a light the same from one shot to the next. It's the great challenge of AI video.
- Seed
- The number that initializes the randomness of a generation. With the same prompt and seed, some models give a very similar result: handy for fixing a detail without losing everything.
- Upscale
- Increasing the resolution of an image or video after generation, for more sharpness.
- Inpainting
- Regenerating a single area of an image (an object, a hand, a face) without touching the rest.
- Lip-sync
- Matching a character's lip movements to a voice.
- Speech synthesis
- A voice generated from text. It carries the narration, and its tone matters as much as the image.
- Credits
- The payment unit of most generation platforms: every image or every second of video uses some up.
- Shot
- A continuous piece of video, between two cuts. A narrative video is built shot by shot.
- Shot list
- The list of shots in a scene, taken from the script: shot size, length, camera movement.
- Storyboard
- The sequence of images showing each shot before you generate it. It saves you paying for useless generations.
- Continuity
- The match between two shots: lighting, position, axis, props. A continuity error shows immediately.
- Art direction
- The visual choices made before the first generation: palette, lighting, era, style. It's what keeps a series together over time.
- Vertical format (9:16)
- The format of short videos watched on phones, on short-video platforms.
- Color grading
- The final adjustment of colors and contrast, so that every shot shares the same light.