Generate two shots of the same scene and the second one forgets the first: a new crowd, a different child in her arms, everyone standing somewhere else. Dawnwood renders each new shot as a continuation of the one before it, so the cut plays as one scene.

Why does the next AI shot not match the last one? A video model renders each clip on its own. It reads your prompt and your reference images, then invents everything else again: the room, the extras, the props in someone's hands, the light, where each person stands. Character references keep faces close, but they do not tell the model what the previous shot looked like at its final frame.
In a short drama that shows up at every cut. Viewers notice a cloak that changes color, a toddler who turns into a baby, a crowd that appears from nowhere. Fixing it by hand means re-rolling until you get lucky, or cutting away so the change is hidden.
A group from our own series, rendered twice from the same prompt and references. On the left it was generated on its own. On the right it was rendered as a continuation of the shot before it. Each side plays the last three seconds of shot one, then the start of shot two.
LEFT: GENERATED ON ITS OWN · RIGHT: CONTINUATION
Every group after the first in a segment continues from the one before it. If a flashback sits in between, the next group continues from the shot before the flashback.
Starts fresh from its references. There is nothing on screen to carry over.
A single re-rendered shot continues from the part of the group just before it, so it lands back in the same staging and light.
Follow the template's camera move instead. The template clip takes the reference slot continuation would use.
Start a Seedance 2.5 series and every shot in a scene picks up where the last one stopped.