October in 8 seconds ๐
Made from a single text prompt. No camera, no people, no editing beyond the text overlay.
Full prompt in the thread ๐
Free Hook Starter for faceless AI video creators: https://t.co/qIGIVOHbo5
@modelstoreai Great test. For breakage it often helps to write the reaction as steps: "the boot lands, thin cracks spread outward from the heel in a spiderweb, the ice stays in place." Saying it doesn't splash or shatter steers models away from the puddle default.
@openclipsai The surprise twin is such a classic. Stating the count often helps: "a single waiter, the only person in the room," repeated at the end of the prompt. Models tend to fill empty space with extra people unless you tell them the room stays empty.
@Dheepanratnam A full 30s from one text prompt is a tough brief. If either model drifts in the second half, rough timestamps usually help ("0-10s: ..., 10-20s: ..., 20-30s: ...") with one action and one camera move each, so the pacing doesn't get left to the model.
@aisamsonreal Keeping the chrome arm and cheek plating through a full turn is impressive, and the time and cost notes are really useful. For a round two, a detail only visible from behind (a cable or patch on the back) would show which model remembers the unseen side.
@Savvy_Erudite@ElevenLabs@ElevenCreative Makes sense, generation keeps getting faster and the edit is where the hours go. One thing that can save time: write the cut list first, with a target length for each shot, so you only generate clips that already have a slot.
@Decoded08 Thanks for this! We tried "subtle natural motion only" on a wide landscape today and it kept the scene almost frozen, so it seems best for figures, where melting limbs are the risk. For wide shots, naming one thing that moves ("fog drifts right to left") worked better for us.
An old green door stands alone in a foggy dawn meadow, ajar, golden sunlight spilling through. Static tripod camera. Breeze blows low fog right to left, grass sways, light glows; door stays fixed. Sound: soft breeze. Free Hook Starter: https://t.co/qIGIVOGDyx
Tip: generate the still first, then animate it. Name what moves (sand, clouds) AND what stays fixed (lighthouse, dunes, horizon). That one line stops the scene from melting.
Free Hook Starter prompts: https://t.co/otRfxsqO2w
A lighthouse 500 miles from the nearest sea ๐๏ธ
So who is it guiding home? One image prompt + one video prompt, no people. Full prompts in the thread ๐
Video prompt (image-to-video): Camera completely still on a tripod, no camera movement, no zoom. Steady desert wind: soft veils of sand haze stream right to left low across the dunes, clouds drift steadily, the lamp glows warm. Lighthouse, dunes and horizon stay fixed.
@openclipsai The vanishing mug is such a classic. Spelling out what happens after the push often fixes it: "the paw pushes the mug off the edge, it tips, falls to the floor and shatters." Models treat the object as disposable unless its fall gets its own beat.
@Everypixelcom Useful numbers. On the refusals: Seedance often trips on a single word (a real name, a weapon, "blood") rather than the whole brief. Swapping that word for a plain description and resubmitting can rescue a lot of them before you switch models.
Tip: if water "slides" like a scrolling texture, take your best first frame and re-run it as image-to-video with that same frame as the LAST frame too. The loop forces ripples to shimmer in place.
Free Hook Starter prompts: https://t.co/fLuagfmWs2
(3/3) Camera completely still on a tripod, locked-off static frame. The only motion is soft rippling caustics shimmering on the walls and ceiling from very gently moving water. Soft deep blue and pale cyan light, slight haze. Sound: room tone, soft hum, faint water lapping.