Guide: connecting shots that flow into each other

trailer-guide.html

@@ -63,7 +63,22 @@
 <p>Music "literally sets the tone and the rhythm"; most trailers use up to three cues, one per act. We do not yet lay music - it is the largest remaining gap - but the cut should be planned for it: a quiet bed under the introduction, a pulse through the escalation, rises into hits in the climax, then the stop. One well-placed hit is powerful; "four is a parody".</p>
 <h2>8. What this means for our pipeline</h2>
 <div class='tw'><table><tr><th>craft rule</th><th>how we do it</th></tr><tr><td>On-camera lines from the book</td><td>checker objects to an on-camera line not in the manuscript; page badges every line</td></tr><tr><td>Narrator</td><td>a cast member with a voice and no appearance; all lines <code>lip_sync: false</code></td></tr><tr><td>Shots run their written length, line placed within</td><td>stretched Action/Cut slate, <code>say_at</code> start / middle / end</td></tr><tr><td>Accelerating climax</td><td>wordless shots may be as short as 1.5 s in the edit (each still costs a 5 s generation)</td></tr><tr><td>One action too big for one take</td><td>split into two shots (the abduction became 17 and 17a)</td></tr><tr><td>Objects and cars stay the same</td><td>tracked props with reference pictures</td></tr><tr><td>The camera can actually see it</td><td>line-of-sight rule (no urn "seen" from the back seat)</td></tr><tr><td>Nothing bought twice</td><td>request-hash guard on pictures, clips and audio</td></tr></table></div>
-<h2>9. A checklist for judging a script before anything is drawn</h2>
+<h2>9. Connecting shots that flow into each other</h2>
+<p>The video model films one shot from one opening picture and one description, with no memory of the shot before. Everything that carries across a cut has to be put in front of it twice.</p>
+<p><strong>The rule: a shot that continues the last one starts on the last one's <em>last frame</em>, and its first sentence says what is done about whatever was left moving.</strong></p>
+<ol>
+<li><strong>The picture.</strong> On the storyboard: <em>Opening picture &rarr; Same as shot NNN's last frame</em> (available once that shot is filmed). That frame has the ball where it actually stopped, the dog's head where it actually turned. A drawn picture would put them where the writer imagined.</li>
+<li><strong>The words.</strong> Open with the state the frame shows and the first thing done about it - the "go get" sentence: <em>"The ball lies where it stopped, six feet inside the arc; Dorothy walks to it, bends and picks it up with her right hand."</em> If it was thrown it is fetched; dropped, picked up; rolled, walked to. Never open with the person already holding the thing the last shot left on the ground.</li>
+</ol>
+<div class='tw'><table><tr><th>carries across the cut</th><th>how to hold it</th></tr>
+<tr><td>the moving object (ball, phone, cup, door)</td><td>last-frame opening + a "go get" sentence; where it is in feet from a fixed thing</td></tr>
+<tr><td>which way people face</td><td>say it - "facing the basket", "her back to camera"</td></tr>
+<tr><td>who is where</td><td>name each person's spot, even the one who does nothing ("Toto still sits beneath the basket")</td></tr>
+<tr><td>the camera side</td><td>"from the same sideline"</td></tr>
+<tr><td>the light and the place</td><td>the scene's background carries them; do not re-describe</td></tr>
+<tr><td>the length of the beat</td><td>one continuous action per shot; a "go get" plus the next beat over ~8 s is two shots</td></tr></table></div>
+<p><strong>Closing pictures are guesses; last frames are facts.</strong> Draw the next shot only after the previous one is filmed, from its last frame; if you refilm a shot, take its last frame again for the shot after it. Thrown to another person: end shot A on the throw, open shot B on the catcher already looking at the incoming ball - never show the flight twice. A hand-off stays inside one shot. If the object is meant to stay behind, say so, or the model brings it along.</p>
+<h2>10. A checklist for judging a script before anything is drawn</h2>
 <ol><li>After the first 15 seconds, am I curious?</li><li>By 35-40 seconds can a stranger say who, what is wrong, what he wants?</li><li>Is every person identified the first time they are seen?</li><li>Does the opposition have a face and a reason by the middle?</li><li>Is every on-camera line the author's? Does each work with no context?</li><li>Does every effect have its cause on screen or in the narration?</li><li>Do the shots get shorter towards the end?</li><li>Does it end on a question, not an answer?</li><li>Would I want to read the book?</li></ol>
 <hr>
 <h2>Sources</h2>