00:00 / 00:00

The prompt-builder skill and first Seedance run

Lesson 619 lessons

Style: 8K IMAX. Photorealistic — no 3D render, no game engine, no game-cutscene aesthetic. Cinematography: Emmanuel Lubezki × Roger Deakins. Lighting: Natural light only — contre-jour backlight, camera on shadow side, atmospheric haze throughout. Key light from sky and windows only. Color: 60:30:10 — dominant / secondary / accent. Camera: Physical cine lens. 180° shutter motion blur. Skin: Pore-level realism — vellus hair, asymmetric moles, capillary flush, pore-shadow matching on-set light. Acting: Hollywood — micro-pauses before reactions, precise eye-line, wet living eyes with catch-lights, visible breath and chest rise. Physics: Gravity and inertia respected — mass has real weight, correct contact shadows. No floating props. Composition: Rule of thirds + golden ratio. Continuity: Characters, props, environment identical across every cut. No identity drift. Technical: 24fps smooth motion. 8K detail. No jitter. Audio: Environmental SFX only. No music. No subtitles. @081d2628-836e-4a30-958c-ac6c893d2eaa — 25, athletic footballer, dark wavy medium-length hair, short beard, dark-olive crew-neck t-shirt and black jeans, reclined in a leather-and-wood lounge chair. 100% matches the reference. @3e9c91c5-fc58-475c-ad0a-a0da2625bc48 — psychologist, dark wavy shoulder-length hair, taupe tailored suit over a cream blouse, seated upright in a rust egg swivel chair with her back to camera. 100% matches the reference. @b8d9698e-25b4-4971-bfee-37368c385665 — warm mid-century-modern therapy room, floor-to-ceiling gridded windows with low sun and foliage outside, terracotta floor, brass arc and table lamps, jade plants, a small tree, a vintage radio on the floor. 100% matches the reference. SCENE CONTEXT: a guarded first beat of a session — she asks her question across the room, then we cut in to read the memory landing on his face. Controlled multishot — strictly 2 shots, one HARD CUT. ACTION Controlled multishot. The only cut is the specified HARD CUT — the camera never cuts by itself. [SHOT 1] (0–9s) — the question Subject: @3e9c91c5-fc58-475c-ad0a-a0da2625bc48 and @081d2628-836e-4a30-958c-ac6c893d2eaa, both seated, present in frame from the first frame. First frame / blocking: @3e9c91c5-fc58-475c-ad0a-a0da2625bc48 fills the foreground lower-center in her rust egg chair, back to camera. @081d2628-836e-4a30-958c-ac6c893d2eaa sits across the room ~4.5m away, screen-right by the windows, reclined in his lounge chair, body three-quarter to camera, facing back toward @3e9c91c5-fc58-475c-ad0a-a0da2625bc48. Both are visible and seated for the entire shot — neither leaves frame, stands, or moves to another seat. Action: ① 0–4s the room breathes — dust in the sun shafts, @081d2628-836e-4a30-958c-ac6c893d2eaa reclined and still, slow chest breath; @3e9c91c5-fc58-475c-ad0a-a0da2625bc48's head barely moves above her chair. ② 4–5s pre-line, @3e9c91c5-fc58-475c-ad0a-a0da2625bc48 takes one quiet breath. ③ 5–9s she speaks, steady and low, "Tell me what you feel when you remember that moment." — voice aimed across the room at @081d2628-836e-4a30-958c-ac6c893d2eaa; he listens without turning his head. Camera: from low and behind @3e9c91c5-fc58-475c-ad0a-a0da2625bc48's egg chair, looking across the room toward @081d2628-836e-4a30-958c-ac6c893d2eaa, her chair large in the foreground. WS 28mm, straight lines stay straight. An EXTREMELY SLOW dolly-in, ~12–15cm total across the 9s — almost imperceptible, physical move, no zoom. Camera on the shadow side, smooth handheld, fine breathing micro-movement only. Style: low sun through the windows as key, ~4800K, contre-jour, foreground egg chair and @3e9c91c5-fc58-475c-ad0a-a0da2625bc48 as a warm near-silhouette, @081d2628-836e-4a30-958c-ac6c893d2eaa with a soft window wrap on one cheek. ⚠️After 9s HARD CUT to SHOT 2 — no transition, no fade. [SHOT 2] (9–15s) — the memory on his face (hard cut from SHOT 1) Subject: @081d2628-836e-4a30-958c-ac6c893d2eaa only, tight on the face, still seated in his lounge chair. First frame: his face fills the frame from the first frame — no empty establishing frame. Action — the question pulls a buried memory up; he fights it (suppressed pain, eyes do the work): ① 9–10.5s: his gaze unfocuses and drifts off past the lens — a thousand-yard stare, pupils settling as if seeing the moment again, not the room. ② 10.5–12s: the masseter at the cheekbone slowly tightens; one delayed, forceful swallow travels the throat. ③ 12–13.5s: the inner brow pulls up and together (corrugator), the eyes gloss over — a wet film pools on the lower rims and catches the window light, but ⚠️no tear falls. ④ 13.5–14.5s: one slow, deep, controlled inhale — chest visibly rises and falls; the jaw trembles once, then clamps tight. ⑤ 14.5–15s: a single slow blink presses the wet back down; he holds it in, eyes still distant. Camera: CU 100mm (≈18° FOV) — background flattened to soft window bokeh, face natural, no warp; eye-level, on the lounge-chair side. Begins strictly static, then an almost-subliminal slow push of ≤8cm. No drift mid-segment. Operator on the shadow side. Style: same ~4800K; window light wraps one side of his face, the far side falls into soft shadow — chiaroscuro, wet catchlights carry the eyes; not flat front light. PHYSICS: real facial micro-movement, throat and chest motion, skin shows pore-level detail and a faint capillary flush; no exaggeration, no mugging. CAMERA (overall) LENS LOCK SHOT 1 = 28mm wide. LENS LOCK SHOT 2 = 100mm tight CU. Camera on the shadow side throughout. STYLE Naturalistic low-key, warm backlit interior, real 24fps film grain, grounded physical-cinema texture. Lubezki natural-light handheld. Continuity: @081d2628-836e-4a30-958c-ac6c893d2eaa, @3e9c91c5-fc58-475c-ad0a-a0da2625bc48, @b8d9698e-25b4-4971-bfee-37368c385665 identical across the cut. CONSTRAINTS POSITIVE LOCKS: exactly two people, both seated and in frame — @3e9c91c5-fc58-475c-ad0a-a0da2625bc48 in the rust egg chair (back to camera, foreground) and @081d2628-836e-4a30-958c-ac6c893d2eaa reclined in the lounge chair across the room; neither disappears from frame, stands, or sits anywhere else; lips move only for @3e9c91c5-fc58-475c-ad0a-a0da2625bc48's one line in SHOT 1; @081d2628-836e-4a30-958c-ac6c893d2eaa stays silent in both shots; tears pool but never fall; in SHOT 2 his eyes close at the end and stay closed to the final frame; dolly push-in and CU push stay slow, smooth, physical only; windows stay bright, faces carry a clear light pattern; identity and wardrobe identical across the cut; clean crisp image. Strictly 2 shots, one hard cut at 8s, no fades. AUDIO: faint warm ROOM TONE throughout — low air hum, soft creak of the chairs, low birdsong through the glass; @3e9c91c5-fc58-475c-ad0a-a0da2625bc48's single line sits over it in SHOT 1, warm and steady; SHOT 2 is room tone only under his breath — one audible controlled inhale, one soft exhale as his eyes close. Diegetic only, no music. 15 seconds. 21:9.
Recreate

The prompt-builder skill and first Seedance run

Higgsfield Academy

Caption

Writing a full shot-by-shot Seedance prompt by hand — camera position, framing, cuts — is the slow part. A custom Claude skill fixes that: loaded with director hacks, it takes your plain description of a scene and splits it into shots automatically, picking the framing (close-up, medium, wide) so you get real camera-angle changes without knowing the vocabulary yourself. Grab the skill file from the sidebar of this lesson and upload it once in Claude under Customize → Skills → the + button — after that, every scene you describe gets the shot breakdown built in.

Ask for the scene, shot by shot

With the skill loaded, you don't write a Seedance prompt — you describe the scene the way you'd describe it to a cinematographer, and the skill turns that into one.

Copy this asset

Write me a prompt using the skill for the first scene. I want the scene to consist of 2 shots. The first shot is from behind the psychologist, with the camera slowly pushing in, as she asks him a question from the script. As soon as she's asked it, the frame cuts to a close-up of Santiago. We need to show him remembering the moment and his inner pain, through his facial expression and emotion. Then he closes his eyes.

Recreate

Naming the shots yourself — a push-in for the question, a hard cut to a close-up for the reaction — is what gives the skill enough to work with. Tag your characters and location with @ the same way you would in the studio, and the skill folds them straight into the output.

Lock the settings, then hit generate

Copy the skill's output back into Cinema Studio and set the model to Seedance 2.0 in 4K, aspect ratio 16:9, duration 15 seconds — that's the standard combo for a realistic, detailed cinematic shot at the highest quality Higgsfield offers. Because the characters and location in the prompt were saved as Elements, the studio pulls them in automatically; no need to re-attach references by hand.

One more thing locks in at this step that's easy to miss: the voice-over. Seedance generates a unique voice for every character sheet you build, and it matches the voice's tone to how the character looks. Lock in the visuals on the character sheet, and you've locked in the voice too — no separate voice step needed.

Previous Next