Seedance 2.5 prompt guide: writing the long take

What the longer ceiling is for

Nothing about how you write changes between tiers of the family: the same shot grammar — marker and seconds, framing and camera, who is in the picture, what happens, the line, the ambience — feeds all three, and the picture is the only thing that moves with the tier. What the family adds is headroom. A take can be a conversation that plays out while two people walk, a push-in that has time to arrive, a reveal that is allowed to land before the cut. If a scene never needed that, a family with a shorter ceiling does the same job; if it did, this is the reason to be here.

One shot or several

The still above is from The Eagle’s Vow, an episode this site took from script to composite on Wan 3.0 · 720P. Its segments were written the way a Seedance 2.5 Pro segment would be, and the long-take storyboard setting that sets a segment to the longer ceiling is chosen when the storyboard is generated.

Decide first whether the beat is one thing or several. If one person crosses a room, says what she came to say and reaches the door, that is one continuous action: write one shot, one framing, one camera move, and give it the seconds it needs.

[Shot 1] 14s: Tracking shot alongside, medium shot. @Mara walks from the desk to the window, @Jules a step behind her. @Mara (quiet, not turning): I already know what it says. (voice: 🔊Mara) @Jules (low): Then why keep it? (voice: 🔊Jules) Ambient: rain on the glass.

If the beat changes framing, it is several shots, and the seconds of each add up to the segment. Writing them as a chain is not a downgrade: a held shot is right when the action is continuous, and cuts are right when the scene turns.

Pace the lines to the seconds

A long take tempts people to put more in it. The measure that keeps it honest is speech: about three English words to the second, so a line needs roughly a third as many seconds as it has words, and the action around it needs its own time on top. Fill the ceiling with dialogue and there is no room for anyone to move. A take that lets two people trade a line each and then breathe reads as a scene; a take that crams the whole argument reads as a recording.

Seconds are tuned in the storyboard, before anything is generated, and changing them costs nothing. Render the draft on Seedance 2.5 · 480P at 19 credits a second, read the pacing, and only then spend on the tier the viewers will see.

End the take where the next one can start

Each segment opens where the previous one ended: whoever was seated, wherever the hands were, that is the starting state for the next call. So write the last shot of a long take to land on a position you would be happy to continue from — a settled pose, a clear framing, nobody caught mid-step. A take that ends in the middle of a gesture hands the next segment a problem it has to solve in its first second. The longer the take, the more this matters, because there are fewer joins to hide in; which segments the composite joins shows where the joins fall.

References and what to attach

The family takes up to ten images and five voice clips per generation, with about fourteen seconds of voice in total, the same as the Wan steps. A long take that carries several speakers needs its voices attached, and a crowd does not need a sheet for every face. Spend the images on the people the camera will find, and name only them in the shot.

When a shorter take is better

Not every scene wants the ceiling. A reaction needs a couple of seconds, an insert even fewer, and padding them to fill a segment adds seconds nobody asked for. Write the length the beat has, not the length the tier has. The price page and the family page have the per-second numbers; the comparison with the 2.0 family is the one to read when the choice is the ceiling.

FAQ

Updated October 14, 2026