
sref-NVBHGP
Layers of hand-cut coloured paper stacked in a shadow box, each one casting a soft shadow on the next.
Paper & cut · Calm
Preview is 9:16. Pick a shape to put it in the prompt; otherwise your AI asks you.




Real clouds and a few floating objects drifting in a pastel sky, slow and weightless.
Drift most of the clip through a pastel cloud room with the things being remembered or hoped for floating past, cutting back to the speaker once or twice for the quiet turn.
Full scene: the scene fills the frame. Stock speaker from Pexels.
THIS IS A MOODBOARD, NOT A TEMPLATE. The style below is inspiration for the look and feel. Keep its materials, colours, textures and the way it moves, but invent your own scenes, characters and objects for this clip. You are not limited to anything shown in the preview, and nothing in it has to appear. Use illustration and animation to help tell the clip's story. STORY FIRST. Read the whole clip before you design anything. 1. Write the clip's arc in one line: setup, tension, turn, pay-off. Name the one line that is the reveal. 2. Tag every sentence with its job: setup, context, tension, turn, reveal, example, number, contrast, list, quote, feeling, metaphor, callback, pay-off, aside. 3. The visuals carry the whole clip. If the clip shows the speaker's face, open on them (the first scene keeps them on screen) unless the person says otherwise. After that, never cut back to the speaker full-frame unless the person asks for it; if they do, only on the ending, never in the middle of an explanation. The speaker can still sit inside a scene (a corner, a split or drawn in) when the person wants to be seen. 4. Build one flowing visual story, not a new picture per sentence. Each scene covers a whole thought (one to three sentences), holds 3 to 8 seconds and turns into the next. 5. Start each scene on the first word of its sentence, or up to half a second before. Never cut mid-sentence. 6. Inside a scene, bring small details in on the words they show (an item, a tag, a tick). Earlier details stay. 7. Choose one or two recurring objects or characters in the first scene and bring them back changed. Callbacks beat new pictures. 8. Build up the same picture where you can (add to it, move the camera across it) instead of replacing it. 9. Energy: about 80 percent still, drifting or gentle. Allow one big move in the whole clip, on the reveal, for 1 to 2 seconds, then settle and hold. 10. Numbers sit on a scale (gauge, bar, table, timeline) and hold long enough to read. Contrasts are a still split or the same layout twice. 11. Feelings get a quiet image, or one of the cuts back to the speaker; never effects. 12. On-screen text names, labels or quotes; it never repeats the caption. Five words or fewer per label. 13. End every scene by settling or thinning out, or by turning into the next scene. No exit effects. 14. The style sets the look. These rules set when, how long and how calmly. If the style's own motion rules ask for more energy than this, use less. PICTURES: this style uses photo cut-outs. Before you start, ask the person one short question: use stills from their own clip, Slice's stock search, their own uploads, or (only if you can make images) images you generate. If they don't mind, use stills from the clip, then stock. VISUAL STYLE: Cloud room Slice style ref: sref-854ZXH LOOK: a cloud room. A soft pastel sky graded from pale blue #CFE3EA through warm white #F7F3EC to peach #F6D9C4. Real photographed clouds, cut out with soft edges, float in three layers, and a few real objects float among them in their natural colour, lightly washed toward peach so they share the same air: a key, a cup, a window, a pale planet as a daytime moon. PICTURES: cumulus clouds against clear sky (to cut out), and quiet everyday objects on plain grounds: windows, doors, keys, cups, chairs, a planet or moon from NASA. Where to get them: public-domain and CC0 collections (The Met Open Access, Smithsonian Open Access, Biodiversity Heritage Library plates, NASA, Wikimedia Commons files marked public domain or CC0), Slice's stock search, or the person's own photos. Cut every object out cleanly; never use a picture with a real brand or logo on it. LINE AND FILL: no outlines and no hard shadows; nothing touches the ground. A faint haze and film grain over everything. CAMERA: a slow, continuous upward drift with parallax: far layers move least, near clouds most. The one peak is a scale reveal: one object drifts to the centre and grows large while a near cloud passes in front of it. MOTION: weightless. Objects bob and turn very slowly, clouds slide sideways. Nothing snaps or bounces. TYPE: none, or one short line in a soft serif if the clip needs it. THE RULE: everything floats; nothing ever lands. DON'T: no saturated colours, no harsh sun, no people, no fast moves, no purple. Treat these as the look, never a template: choose fresh objects, words and compositions from what this clip says. Accent: as in the palette. Frame shape: ask me before you start (9:16 vertical, 4:5 portrait, 1:1 square or 16:9 wide). Energy: normal. TECH: one HTML file, one GSAP 3 timeline created paused as window.__timelines.main, duration = the scene's length, transforms and opacity only.
For AIs: this style as plain text.
Need the clip first? Shorts by Slice cuts your long episode into captioned clips. Then give one this look.
Try Shorts by SliceShorts by Slice cuts your long episode into captioned clips. With Slice connected, your AI can then build scenes for a clip in Cloud room.
Shorts by Slice
From the team behind these free tools
One video. A week of posts.
Upload an episode and get back the moments worth posting, cut and captioned.