Why shots lose their spatial logic
The most common failure isn't ugly visuals — it's the audience noticing right away that someone isn't where they were. In shot one, the lead is standing to the right of the table with a voice recorder near her hand. In shot two, now a close-up, she seems to have jumped to the left. In shot three, the other character's shoulder shows up facing the wrong way. When creators review clips one at a time, they tend to focus on faces and lighting and miss that the scene no longer cuts together as one space.
Blocking consistency in AI short drama starts by answering: who stands where, who's looking at whom, how close is each person to which prop, and can the camera actually cut to the other side. It's not an unnecessary storyboarding lecture — it's the foundation that saves rework. Promises of one-click video generation wear thin fast; what's actually useful is knowing which step cuts down on redoing shots.
You can build your characters in the creative workspace, lock identity with a character turnaround sheet, and break down shots with the AI short drama guide. The staging diagram sits in between, pinning down the spatial relationships.
Rules for staging that holds
Blocking is a story relationship first. Who's near the door, who's blocking the way, who's closest to the evidence — the audience reads all of it as power and emotion. Don't just write "office argument." Write "front-right, hovering over the voice recorder" and "back-left, near the door."
Solo blocking needs weight distribution too. A lot of solo stances fall apart because every shot looks like a passport photo — feet evenly planted, shoulders level, staring straight into the lens.
Multi-character blocking needs an axis. If the camera crosses to the other side of a table where two people are talking, their left-right relationship flips.
Props are spatial anchors. A contract, a phone, a voice recorder, a door handle — anything that matters to the story needs a fixed position. When a prop jumps, the blocking falls apart with it.
Blocking doesn't lock down the performance. It locks the spatial relationships — expression, pacing, and small gestures can still vary.
Blocking also serves the edit. Knowing ahead of time which shot is the relationship shot and which is the reaction shot gives you somewhere to hide a slightly-off lip sync later.
Stable blocking is what lets an edit decide what to keep and what to cut.
How to build blocking consistency: from goal to checklist
Step one: state the spatial goal — who's pressuring whom, who wants to leave, who's protecting the evidence. Without a goal, a staging diagram is just a scatter of dots.
Step two: sketch a top-down view in writing. For example: conference table centered, door back-left, window back-right, voice recorder just right of center on the table. ZH-01 stands front-right, ZH-02 sits back-left.
Step three: give characters codes. Codes work better than names in a shot list — ZH-01 is always front-right, facing back-left; ZH-02 is always back-left, facing front-right.
Step four: mark camera positions. Camera A covers the relationship shot, B covers the close-up, C covers the reaction, D covers the voice recorder close-up. Specify which positions can't cross the axis.
Step five: write the sightlines. Looking at the other character, looking at the voice recorder, looking at the door — spell it all out. Blocking and sightlines together are what tell the audience everyone's in the same scene.
Step six: leave room for subtitles. Vertical short dramas usually put subtitles in the bottom third — don't let a mouth or a key prop end up covered.
Step seven: review before you polish. Check the left-right relationship, foreground/background, prop positions, and sightlines first — faces and image quality come after.
Where blocking lives in your project
Blocking isn't a standalone diagram — it lives in the shot description, alongside the character assets (face, outfit, voice), the shot list in Storyboard, and the voiceover in Edit, all under the same project. That way, when a shot breaks, you can tell whether it's the face (the asset wasn't referenced correctly), the blocking (the shot description wasn't specific enough), the lip sync (the voiceover wasn't bound to the right character), or a prop position that failed.
The minimal structure: the My Assets answers "who is this person," the staging diagram answers "where are they," the shot list answers "what happens in this shot," and the review checklist answers "is this cuttable." Use the AI short drama guide to break down episodes, and save an approved staging version in the creative workspace.
Here is how to wire this up inside a single ArcLoop project.
Open your IP Project and go to My Assets. Create a character asset for each person in the scene — give each one a name, then go to Library, upload your reference images, and Bind the clearest full-body image as the Main image. Add any additional angle or costume shots as Reference images. Do the same for any scene asset (the conference room, the rooftop) that appears in more than one episode.
Once your assets are ready, open the Episode and click Generate Shots. The AI breaks the episode into shot cards. In each shot card's description, type @ and select the character or scene asset — this is what holds the spatial identity across cuts, not a repeated written description of the face or outfit.
Write the blocking directly in the shot description: camera position, where each character stands relative to the props, and which direction they're looking. One action per shot. For example: @Liang Yu'an stands front-right, hand near the voice recorder. @Cen Luo is back-left near the door, holding the contract. Camera A, axis held.
Generate the image for each shot first and confirm that left-right positions, prop placements, and sightlines match your staging notes before you generate video. Use the AI Chat Panel to batch the video step: Generate videos for Shots 1, 2, and 3. If one shot breaks, regenerate only that card — the asset references keep the others stable.
Example 1: solo stance diagram
Medium shot of @Shen Zhihe on the rooftop at night. She stands with her left foot forward bearing her weight, body turned slightly right, left hand gripping the strap of her canvas bag, right hand raised holding the voice recorder toward the door at frame-left. Wind pulls at her cream-gray trench coat. The railing is behind her. Camera is level, slightly low angle. Cold ambient light from the city below; no additional fill. Her eyes are on the door — not the camera.
Example 2: two-character confrontation staging sheet
Wide shot of @Liang Yu'an and @Cen Luo in the conference room. @Liang Yu'an stands front-right, facing back-left, right hand resting near the voice recorder on the table. @Cen Luo stands back-left near the door, facing front-right, holding the old contract in his right hand. Long table centered in frame; door visible back-left, window back-right. Camera A — relationship shot, axis held. Neither character moves from position; the contract stays in @Cen Luo's right hand and the voice recorder stays on the table front-right. Overhead fluorescent light, cold and flat.
Example 3: blocking review and minimal rework
Review Shot 2 for blocking consistency against Shot 1. In Shot 1, @Liang Yu'an is front-right and @Cen Luo is back-left near the door. Check: left-right relationship preserved, foreground and background positions unchanged, voice recorder still on table front-right, contract still in @Cen Luo's right hand, sightlines still crossing correctly, axis not crossed, bottom-third subtitle zone clear of mouths and key props. If the left-right relationship has flipped or a prop has changed hands, mark as redo. If lip sync is slightly off, mark as minor fix — cover with the reaction shot of @Liang Yu'an.
The mistakes that break a scene
Mistake one: describing only the character, not the space. However stable the character looks, they'll seem to teleport into a new room every time the camera changes.
Mistake two: treating the staging diagram as art. It doesn't need to look good — it just needs to communicate left-right, foreground-background, props, and camera positions.
Mistake three: every solo shot facing straight at the camera. Characters in a short drama should be looking at the door, at a phone, at the other character, at the evidence — not posing for a selfie in every shot.
Mistake four: cramming too many people into one frame. Three characters delivering long lines at once drags down blocking, lip sync, and sightlines all at the same time. Split it into reaction shots whenever you can.
Mistake five: force-cutting a shot where the blocking has already broken. A reversed spatial relationship usually can't be salvaged.
FAQ
Does solo blocking in AI short drama need a diagram too?
At minimum, yes. Write down foot position, weight distribution, hand gestures, sightline, and any key props.
Does the staging diagram have to be drawn out?
No. A written top-down description is enough — where the door is, where the table is, where the characters are, where the cameras are. What matters is that every later shot references the same version.
What's the most common mistake in two-character dialogue?
Left-right relationships and sightlines. If one character looks left and so does the other, the audience reads it as two people not actually talking to each other.
Does blocking limit performance?
No. Blocking locks the spatial relationships — performance still comes through expression, pacing, hand gestures, and voiceover delivery.
In ArcLoop, should you build blocking or generate video first?
Build the character assets and blocking first, then generate video. Otherwise, when something breaks, you won't know which layer actually failed.
