Timelapse is one of the most searched formats in AI video — and one of the most technically demanding to write correctly. The default prompt fails in a predictable way: it names the subject and the time scale ("show a city growing over 50 years"), and the model produces either a static image with movement artifacts or a video that changes too uniformly, with no sense of era, phase, or structural transformation.
What separates strong Seedance timelapse prompts from weak ones is not the subject — it is the structural architecture. The best prompts do three things that generic prompts omit: they anchor the camera (fixed, locked, same angle throughout, so the transformation is measurable against a constant frame); they label phases (Phase 1, Phase 2, or named eras) so the model has a structure to sequence rather than a fog of change to navigate; and they specify the day-night cycle as a pacing mechanism, because the rhythm of light change is what makes a timelapse feel like time passing rather than things randomly changing.
The five prompts below cover five distinct timelapse contexts: a civilization-scale dolly shot that moves forward through 10,000 years without cutting; an aerial locked-angle record of Dubai's transformation from desert to megacity; a static documentary timelapse of a rural road rebuilt in five phases; a fixed-angle witness to one gate across all twelve months; and a conceptual hyperlapse that trusts the model to interpret evolutionary sequence from a short label. Together they illustrate the core grammar of timelapse prompting in Seedance.
1. The civilisation dolly — one continuous forward shot through 10,000 years
See the full prompt on scenic.sh →
"One continuous forward dolly — no cuts. Landscape transforms around a steadily advancing camera as if walking through 10,000 years."
Why this works: At 81 likes, this is the highest-rated timelapse prompt in the Scenic gallery, and its core structural innovation is the continuous dolly as a temporal anchor. Rather than cutting between eras, it advances the camera forward at a constant pace and uses the landscape on either side to transform through successive civilisations: river valley mud-brick → stone fortifications → African empires with laterite soil → Andean stone terraces → industrial cobblestone and smokestacks → modern glass towers → launch pad and black space. The continuity rule is explicit: "No cuts." Each era bridges to the next through shared shapes — "walls become walls, roads become roads" — so the camera never needs to jump; it finds the shape that carries forward into the next period. The color grade is its own timeline: "Ochre to blood-red to emerald to gold to steel-blue, each era bleeding into the next," which means the model has a visual temperature arc running parallel to the architectural transformation. The lighting ages with civilisation: "Golden hour shifting to overcast to firelight to electric — the light itself ages." The camera behavior is locked: "Constant slow forward dolly at chest height. No lateral movement, no tilt." The final beat — "Open concrete ahead. A rocket on a launch pad. Ignition. The rocket rises out of frame. Camera advances into dissipating exhaust revealing black sky and stars" — resolves 10,000 years with a single destination.
The takeaway: use a continuous dolly as a temporal thread — a camera that never cuts, never jumps, never tilts gives the viewer a single unbroken perspective on transformation. Bridge eras through shape continuity ("walls become walls, roads become roads") so each transition has a physical handoff rather than a dissolve. Assign a color-temperature arc that runs parallel to the architectural arc; the light aging with civilisation is as much of the time-passage signal as the architecture changing.
2. The city growth aerial — locked angle and the Dubai desert-to-megacity arc
See the full prompt on scenic.sh →
"Camera maintains the same aerial angle throughout so the transformation is continuous and unbroken. Photorealistic, IMAX cinematic quality."
Why this works: At 38 likes, this prompt's defining technique is the locked aerial angle as a transformation credibility system. The camera position is established in the first sentence — "wide aerial shot of pure empty desert — golden sand dunes, nothing visible to the horizon, harsh midday sun, 1970s" — and never moves. Every subsequent change is measured against that fixed frame: a cluster of low buildings near the coastline, roads cutting through sand, the first skyscrapers at small scale then taller then enormous, the Palm Jumeirah forming in the water, then the Burj Khalifa "shooting above everything." The locked angle is not a passive choice — it is what makes the transformation legible. If the camera drifted or re-framed, the viewer would lose the reference point and the scale of change would become unmeasurable. The day-night cycle does double structural work: "Day and night cycling rapidly — blazing golden days, then nights with thousands of lights reflecting off the Persian Gulf, the city glittering from horizon to horizon." The night frame introduces a second visual system (city lights, Persian Gulf reflection) that the desert frame lacks entirely; the contrast between desert-by-day and megacity-by-night is the single most dramatic transformation the prompt can show. The final camera move — a pull-back revealing "the full Dubai skyline we recognize today — Burj Khalifa, the Palm, the coastline, all lit up at night" — is the only movement permitted because it is a reveal, not a reframe.
The takeaway: lock the aerial angle for the entire duration — the transformation is only legible when every change is measured against a fixed frame; a drifting camera makes scale change invisible. Use the day-night cycle to introduce a second visual system (city lights, surface reflections) that the empty-landscape opening lacks, so the contrast itself becomes part of the transformation arc. Reserve any camera movement for the closing reveal, not for reframing during the sequence.
3. The road reconstruction — phase labeling and documentary locked-tripod technique
See the full prompt on scenic.sh →
"Static locked camera, wide shot, timelapse, same exact angle throughout entire video. Starting scene: muddy rural village road with deep tire tracks and puddles."
Why this works: At 32 likes, this prompt operates as a five-phase documentary timelapse with the same structural grammar as a construction schedule. Phase 1 is earthwork (excavator levels the mud road, grader smooths the surface, drainage pipes installed along the sides). Phase 2 is base layer (dump trucks pour gravel, road roller compacts layer by layer). Phase 3 is asphalt (paver machine moves slowly laying hot black asphalt, steam rising, heavy roller compacting smooth). Phase 4 is finishing (workers paint road markings, install curb stones, replace street lamps with LED poles). Phase 5 is landscaping (grass seed planted, young trees planted, wooden fences repainted, "greenery grows rapidly in timelapse"). Each phase has named equipment, named materials, and named workers — not because Seedance needs every detail, but because the specificity gives the model a structured sequence to order rather than a pool of construction imagery to randomly select from. The day-night-day cycle is specified as atmosphere: "natural lighting changes from morning to evening through the process, day-night-day cycle visible" — time-passage signaling without requiring a separate lighting section for each phase. The final frame is a deliberate bookend: "Same camera angle: clean smooth asphalt road, green lawns, renovated houses, bright street lights on, warm golden hour lighting" — which explicitly mirrors the opening frame and makes the before-after legible without a split-screen.
The takeaway: use numbered phase labels to structure construction timelapse — Phase 1, Phase 2, Phase 3 are scheduling grammar that the model can sequence; a prose description of "construction happening" produces random order, but labeled phases produce a directed arc. Specify named equipment per phase (excavator, grader, paver, roller) rather than generic "machines"; equipment specificity tells the model which visual register each phase belongs to. Close with a final-frame description that explicitly mirrors the opening frame so the before-after is a built-in bookend.
4. The twelve-month temple — static witness and the full seasonal cycle
See the full prompt on scenic.sh →
"Camera never moves. The gate never moves. Everything around it transforms."
Why this works: At 5 likes, this prompt contains the most technically specific structural concept in this selection: the "static witness" — a fixed subject and fixed camera that hold identity across twelve complete seasonal transformations. The Kaminarimon Gate of Senso-ji is the anchor. It appears in every frame. The camera is locked to the same ground-level angle, same framing, same composition for the entire year. The only variable is everything else: snow falling on the great lantern (January), bare trees and New Year decorations (February), cherry blossom buds (March), full pink explosion with petals drifting across the stone path (April), paper lanterns strung across the approach for summer festival (July), autumn maple fire (October–November), first snow returning (December). The prompt lists all twelve months explicitly, each with distinct atmospheric detail — "rain" for June, "heat haze" for August, "typhoon season energy" for September. This specificity matters because months without explicit direction tend to collapse into their season's visual average; "cold clear blue sky, bare trees, New Year decorations still hanging" gives February a unique fingerprint that prevents it from reading as identical to December or January. The sound design is specified as a parallel transformation track: "winter silence and soft snow, spring breeze and petals, summer festival drums and crowds, autumn wind through dry leaves, back to winter silence" — the audio arc mirrors the visual arc and returns to its starting state, completing the cycle.
The takeaway: use "static witness" architecture for seasonal timelapse — a fixed subject and fixed camera make every change visible as change relative to an unchanging anchor; movement in any single month becomes legible because twelve months of non-movement surround it. Specify every month with a unique atmospheric fingerprint (not just season labels); months without distinct detail collapse into their season's average. Build the audio arc as a parallel track that mirrors the visual arc and returns to its starting state to signal the cycle completing.
5. The evolution hyperlapse — conceptual brevity and trusting the model's visual library
See the full prompt on scenic.sh →
"human evolution over time a chronological hyper lapse video seamless transitions"
Why this works: At 6 likes, this prompt is the opposite of the four above — it is 12 words, no camera direction, no phase labels, no color grade, no equipment names. Its technique is conceptual brevity: the phrase "human evolution over time" points at a subject where the model has extremely strong visual priors from training data. Human evolutionary sequence — australopithecine through homo habilis through homo erectus through archaic and then modern homo sapiens — is one of the most heavily illustrated scientific progressions in human publishing. The model carries this library internally. Adding phase labels or equipment names would not improve on what the model already knows about the sequence; it would only constrain the visual execution in ways that might omit what the model would have shown anyway. "Chronological" does the necessary ordering work, and "seamless transitions" specifies the only technical behavior that matters (no hard cuts between evolutionary stages). This prompt works as a case study in when NOT to over-specify: if the model's existing visual knowledge of a subject is richer than your ability to describe it, a brief accurate label outperforms a detailed description that inevitably misses what the model would have included.
The takeaway: use conceptual brevity when the model has strong visual priors on the subject — evolutionary sequence, geological timescale, and historical transformation are domains with extensive illustration in training data; a precise short label captures everything the model knows about the subject. "Chronological" and "seamless transitions" are the only structural constraints needed when the sequence is already culturally canonized. Reserve detailed phase labeling (as in prompts 3 and 4 above) for transformations that are domain-specific or visually non-canonical — road construction phases and twelve-month temple seasons are not in the model's automatic visual library the way evolutionary sequence is.
Seedance timelapse prompt cheat sheet
Across all five, the structural techniques that make Seedance timelapse prompts work:
- Lock the camera for the entire duration — fixed tripod, locked aerial angle, or steady dolly with no lateral movement gives the viewer a constant reference point against which every change is measurable; a moving camera makes scale change invisible.
- Use phase labels to structure transformation — numbered phases or named eras (Phase 1 / Phase 2, or era labels by civilisation/month/decade) give the model a sequence to order rather than a pool of imagery to randomly select from.
- Specify the day-night cycle as a pacing mechanism — time-passage in timelapse is signaled by light cycling from morning through evening through night and back; naming the cycle explicitly ("day-night-day cycle visible") makes the model run it as a pacing system, not as random lighting variation.
- Use shape continuity to bridge eras — "walls become walls, roads become roads" is a transition grammar that lets a single-take dolly move through centuries without a cut; find the shared shape between the ending state of one era and the starting state of the next.
- Apply conceptual brevity when model priors are strong — subjects with rich illustration in training data (evolutionary sequence, famous city growth, canonical seasonal cycles) work with brief accurate labels; save detailed specification for domain-specific transformations the model cannot reconstruct from a label alone.
Browse the Scenic documentary gallery for more transformation and time-passage prompt examples, or check the Scenic nature gallery for seasonal change techniques. Read how to write Seedance 2 prompts for the complete cinematic prompting guide.
