Office Mishap AI
An 18-second AI office mockumentary built from three reference images. The chair tips, nobody reacts, one deadpan line lands. Remix it with your cast.
# office-comedy# ai-drama# mockumentary
Domo_1799028
Office Mishap AI: A Mockumentary Scene You Can Remix
This is an 18-second office comedy shot like a single-camera documentary — a manager tips his chair back one degree too far while explaining something nobody asked about, and his colleague watches him go without lifting a hand. A DomoAI user made it in a single generation from three reference images: one for him, one for her, one for the office. No camera, no crew, no set.
The joke is in the restraint. Nobody plays it for laughs, nobody gasps, and the whole thing lands on one flat line delivered straight to camera four seconds later. That deadpan register is the reusable part — and because this is a template, you can keep the format, the timing and the four-shot structure, and put your own cast in the chair.
How to make your own AI office mockumentary
STEP 1
Hit Remix and swap the references
Remix opens this scene in Omni Reference with everything already in place: the three reference images and the full prompt, timings and all. Each reference has one job — Image 1 and Image 2 control the two characters' faces, build and wardrobe, and Image 3 controls the office itself. Replace the people with your own and keep the room, or keep the cast and move them somewhere new. Use your own photos, your own original characters, or images you have the rights to.
STEP 2
Change the incident, keep the structure
The bones are worth keeping: 18 seconds, four shots, hard cuts at 5s, 10s and 14s, and a talking-head at the end that says the quiet part out loud. Swap what goes wrong and rewrite the two lines of dialogue, but give every beat one camera angle and one action, and save the last shot for the payoff line. A timestamped structure holds together far better than a paragraph describing everything at once.
STEP 3
Generate with Audio on
Assign each line to a named character with a timestamp, describe the sound the room actually makes — keyboards, a chair creak, one impact off-frame — and use no music. Play it straight: no gasp, no lunge, nobody laughing on camera. Check faces and lip movement first, then keep the references that worked so your cast is the same person in the next episode.
How This Scene Was Made
Omni Reference
Workflow
3
Reference images
18s
Four shots in one pass
9:16 · 720P
Vertical, with native audio
Frequently Asked Questions
Can I put my own characters in this office scene?
Yes — that is what the template is for. The four-shot structure, the timing and the comic register are inherited; the cast is yours to replace. Swap the reference image for each character and the same scene plays out with your people in it. Use your own photos, your own original characters, or images you have the rights to.
Are the spoken lines part of the same generation?
Yes. Both lines in this scene were written into the prompt with a named speaker and a timestamp, then generated together with the video using Audio on. There is no separate voiceover or dubbing pass, and no subtitles are burned into the frame.
Can I move the scene out of an office?
Yes. The third reference image is the location — it controls the cubicle geometry, the doorway, the ceiling lights and the direction the daylight comes from. Replace it with a kitchen, a classroom or a shop floor and keep the two characters, and the same joke structure carries over. Change one of the three references at a time so you can tell what each one is doing.
How long can my version be?
Omni Reference generates from 4 to 30 seconds. This scene uses 18, split across four shots. Longer is not automatically better for a short drama beat — the reason 18 works here is that every second is doing something, and the cut points are what carry the comedy.
What resolution and aspect ratio does it output?
This scene is 9:16 vertical at 720P, which is what makes it postable to short-form feeds without reframing. Omni Reference outputs at 480P or 720P and supports 9:16, 16:9, 1:1, 4:3 and 3:4, so you can pick the ratio your platform wants before you generate.
Why does this look like a real documentary rather than AI video?
Because the prompt asks for the flaws a real camera has. It specifies broadcast documentary texture with slightly flat contrast, a neutral-to-green office white balance, mild noise in the shadows, visible skin pores, no digital sharpening and no beauty smoothing. The camera holds a constant small handheld correction and the autofocus hunts once. Those imperfections are what read as real, and leaving them out is what makes a clip look generated.
Do I need to film anything or hire actors?
No. Everything here came from three still images and a written prompt — no camera, no crew, no location booking. The performances, the chair physics, the room tone and the dialogue were all generated in one pass.
How do I turn one scene into a series?
Keep the reference images from a take you liked and reuse them for the next scene — same cast, new situation. That is what holds faces consistent from episode to episode, which is the hardest part of a vertical short drama. Number your episodes and end each one before the situation fully resolves.