> ## Documentation Index > Fetch the complete documentation index at: https://www.latentcast.com/docs/llms.txt > Use this file to discover all available pages before exploring further. # Backgrounds > Changing what sits behind the presenter: settings that apply instantly, new images and clips that ask first, and how to write a motion prompt. Every scene has one or more background layers behind the presenter. Ask the [tuning chat](/docs/guides/stories/tuning-chat) to list them and it will tell you what each scene has and when each layer is on screen. > *"What backgrounds does scene 3 have?"* ## Changes that happen instantly These touch settings, not pictures, so nothing is generated and nothing is charged. * **Hide or show** a layer. * **Blur** it, so the presenter reads more clearly against a busy image. * **Slow motion across a still**, drifting in, out or across. Useful on a photograph that would otherwise sit dead on screen. * **Fade** between one layer and the next. * **Move a boundary** — when a layer starts and ends. Moving a boundary makes its neighbours stretch or shrink to fit. No layer can go below a second and a half, and if your request would push one under that, the chat explains rather than doing something surprising. ## Changes that generate something These show an [approval card](/docs/guides/stories/tuning-chat) first. * **Another take** of the same idea, when the composition is right but the picture is not. * **A new image** from your description. * **A short video clip**, either a new one or motion added to the image already there. * **Your own picture**, used exactly as supplied. ## Image or video? A still image with slow motion across it is cheap, quick, and right for most scenes. A generated video clip earns its cost when the movement is the point: a product in use, a place, an atmosphere that a static frame cannot carry. ## Writing a good motion prompt The single most useful habit when asking for a video background: **Describe what happens in the scene, not the camera.** "Steam rising from a coffee cup, morning light on the table" gives you something alive. "Slow zoom in" gives you a lifeless drift you could have had for free with motion across a still. Animation can only move what the picture already contains. Asking for a person to walk into an empty room will not add them. If you need something that is not there, generate a new image first, then animate that. Generated clips run between 5 and 15 seconds. If a scene needs longer than the clip, the chat will offer to shorten the window or split it across two layers. ## While a clip renders Video generation takes minutes. The preview keeps playing with a placeholder in the slot, and the real clip swaps in when it is ready. You can keep working meanwhile. **If a clip fails, its scene keeps the placeholder and shows a Regenerate action.** A placeholder left in place will appear in the finished video, so clear any failed background before you render. This is deliberate: it makes you look, rather than quietly shipping a gap. ## Deleting, and the safer alternative Hiding a layer takes it off screen and keeps it. Deleting removes it, and its time is given to its neighbours, which changes the timing of the scene. **When in doubt, hide.** You can always delete later; you cannot un-delete. ## Changing several at once Ask for a setting across a whole story — *"blur every background"*, *"add slow motion to all the stills"* — and the chat shows one card covering all of it. One approval, and one undo for the lot. ## Next Change what the presenter says, and what follows from it. On-screen text, where the presenter stands, and scene order.