How to Make an AI Music Video, Shot by Shot

Most guides on how to make an AI music video stop at picking a tool and typing a prompt, so you never see how a whole song's worth of shots is planned, generated and held together. We followed Catch Out, a 4:06 short by Dr. Dreams from the Summer 2026 Flick Residency, with three songs written by Dr. Dreams on its soundtrack, from its first frame at a rail-yard fence to the spotlight that catches its crew. The film was built from 29 video clips and 62 images generated on its public canvas. This guide shows how to make an AI music video in eight steps, shot by shot, each with the real frame and the prompt behind it.

How is a music short laid out against its songs?

How do you make an AI music video, shot by shot?

1. Step 1

   Cast the crew once

   Make a turnaround sheet for every character before the first shot, front and both profiles on white, so every video prompt can cite the same face and clothes for the length of the song.

   • Dr. Dreams made sheets for Casey, Patches, Gumby, the railroad officer called the Bull, and Thor the dog.
   • Casey gets a second sheet with a buzzed head, because his look changes at the campfire and 8 of the 29 video prompts call for the buzzed head.

   

   Professional character turnaround reference sheet. Show the SAME man (Casey) in 3 panels side by side: front view, left profile, right profile. Full body, neutral A-pose, plain white seamless background, even studio lighting, orthographic framing. Identical face (lean, 23, long unkempt brown hair, light stubble, wide earnest eyes), same costume in every panel: worn blue/brown plaid flannel over black tee, ripped dirty jeans, heavy boots. Label each panel: FRONT, LEFT, RIGHT.

2. Step 2

   Plate every place

   Write each location as a headed note and generate it as an empty plate, so every stop of the song has a fixed place before any character walks into it.

   • The canvas holds 8 notes headed "Montage", from a store dumpster to a highway overpass in the rain.
   • Each plate is empty on purpose, "framed to leave foreground room for figures", so the crew can be placed in it later.

   

   # Montage — Truck Stop at Night
   
   A truck stop parking lot at night, rows of parked semis with running lights, tall sodium lamp poles casting orange pools over wet asphalt, a lit fuel canopy and diner sign glowing in the background. Empty scene, no people or animals, framed to leave foreground room for figures sitting and splitting food. Moody cinematic photography, moody night color, orange and cyan light mix.

3. Step 3

   Build the opening frame

   Compose the first shot as a still with every character in place, then animate from it, so the song opens on a frame you chose rather than one the video model guessed.

   • The prompt numbers its reference images, "Image 1-3 are Casey", so the model knows which face belongs to whom.
   • The finished film opens at this fence, with the crew hiding from a rail-yard spotlight.

   

   Build a cinematic night scene. Image 1-3 are Casey (jittery 23, long unkempt hair, anxious eyes). Images 4-6 are Patches (25, sun-scorched, leather jacket, calm). Images 7-9 are Gumby (23, tall lanky, red bandanna, thrift clothes). Image 10 is Thor the floppy-eared hound. Place all four crouched low together in tall weeds by a chain-link fence at night, a distant rail-yard spotlight raking behind them. Thor asleep across Casey's boot. Faces half-lit, breath visible, cold blue moonlight, shallow depth of field, 35mm film grain, tense stillness.

4. Step 4

   Time the action in beats

   Write a long action as timed beats inside one clip, such as [00:00], [00:04] and [00:08], so the run, the jump and the landing play as one take you can lay across a stretch of the song.

   • This train hop is five beats in one clip, made with Seedance 2.0 from a still of the crew beside the boxcar.
   • The prompt names the side door four times and rules out the end of the car twice, pinning down a detail a video model easily gets wrong.

   

   

   00:15

   A continuous 15-second night sequence, strict profile perspective throughout, camera parallel to the moving train. [00:00] Close-up tracking shot on their running feet pounding across the gravel, past the train wheels. [00:04] Profile shot tracking alongside the three punks and their dog as they run parallel to the boxcar, keeping pace with the wide open SIDE sliding door in the middle of the car's long side wall; the train is moving quickly and picking up speed. [00:08] Medium profile shot on Patches (woman in the leather jacket) as she runs alongside the open side door, then grabs the edge and swings up sideways through the open side door into the dark interior, entering through the side of the car, never the end. [00:11] Wide profile tracking shot as Gumby (tall, red bandanna) scoops up Thor the dog while running and hands the dog up sideways through the open side door to Patches, who grabs him. [00:14] Close shot as Patches sets Thor down safely on the boxcar floor inside. Cold blue moonlight, distant yard lights, film grain, kinetic handheld urgency, breath visible in the cold air. They always enter through the open door on the side of the boxcar, not the back end.

5. Step 5

   Give the music a quiet shot

   After a loud passage, give the song a slow shot with almost no action, so the music carries the moment and the audience gets a breath before the next run of cuts.

   • In the cut, the first night in the boxcar sits between the train hop and the road montage.
   • The beats are small on purpose: sit down, hang your legs over the edge, pass a bottle, smile.

   00:13

   interior moving box car Empty Boxcar Interior, Rear Door Open. live action film. 
   [00:00] The characters Casey, Road-Worn - Character Ref SheetPatches with Stick-and-Poke TattoosGumby Character ReferenceThor Character Reference jog into the scene from behind the camera and sit on the floor of the box car, exhausted. breathing heavy. camera dollies. 
   [00:03] close on Casey as he sits at the edge of the open door, his legs hanging over the edge. 
   [00:05] over casey's shoulder, looking out the moving fields beyond. 
   [00:08] cut to camera push in close on patches who takes a swig from a bottle. she tosses it over to gumby. 
   [00:10] cut to gumby who catches the bottle. drinks from it. Thor sits next to him. 
   gumby. 
   [00:13] cut back to three shot of all of them. smiling. happy.  camera tilts up toward ceiling. 

6. Step 6

   Cut the montage one stop at a time

   Give each stretch of the montage one place and one small action, so every cut on the beat lands somewhere new and a long road fits into about a minute of film.

   • Catch Out's road montage runs from about 0:40 to 1:52 and visits six places, one after another.
   • This truck stop shot cites the store plate and the Patches sheet by link, and its whole action is Patches chatting with the clerk while Gumby walks out.

   00:04

   live action film. [00:00] interior gas station Flying J Truck Stop Store cut to the checkout counter where  [00:09 handheld camera tracking gumby as he walks to the exit. he walks past Patches Patches Character Turnaroundpatches is chatting and flirting with the teenage clerk behind the counter. camera handheld

7. Step 7

   Make the props

   Generate any object the story turns on as its own product shot, so it looks the same in every hand and every cut of the song.

   • The campfire scene turns on stolen clippers: Patches buzzes Casey's head, and he keeps the buzz cut for the rest of the film.
   • The canvas holds three clipper props, and this cordless one rules out a power cord three times in one prompt.

   

   # Prop: Brand New Cordless Hair Clippers
   
   A brand new cordless consumer electric hair clipper, matte black plastic body with a polished chrome adjustable guard and a crisp sharp trimming blade, pristine and unused straight out of the box. Cordless with NO power cord and no cable of any kind. Single object presented as a clean studio product shot, centered on a pure seamless white background, bright even soft studio lighting, sharp focus, subtle soft shadow beneath, no hands, no people, no packaging, no cord. Crisp commercial product photography.

8. Step 8

   Stage the turn

   Write the story's turn as one clip that escalates beat by beat, from a run to a look back to a blinding spotlight, so the picture builds to its peak the way a song builds to its last chorus.

   • At about 2:48 in the film, the crew freezes in the glare of the security truck's spotlight.
   • The prompt ends each beat on what the audience should feel, "hard tension, engine roar".

   00:15

   ive action film. no background music. night, trainyard Trainyard Edge, No Fences. [00:00] camera dolly, tracking along side the crew Patches Character TurnaroundGumby Character TurnaroundCasey - Turnaround Sheet, Buzzed HeadThor Character Turnaround as they run across the tracks [00:03] cut to close on casey, glancing over his shoulder as he runs, camera handheld
   [00:06] cut to casey's POV looking back as a  second security truck suddenly pulls out onto the road  behind him, headlights and a blinding white spotlight blasting straight at him [00:09] camera tracking behind the three of them running as a security truck Railroad Security Truck suddenly pulls out in front of them, and screeches to a stop and shins a spotlight on them. the three young people freeze, throwing hands up against the glare, silhouetted, caught. hard tension, engine roar, no music.
   [00:12] cut to medium camera orbits in front of the three of them, they're blasted with the spotlight, squinting against the harsh light. 

   Open the Catch Out workflow

What did one music short take to make?

What holds an AI music video together?

1. 1. Lock the faces first

   Name the reference sheets in the prompt and ask for a match, as the train hop shot does, so a cut on the beat never swaps one face for another.

   Keep Patches, Gumby, and Casey exactly matching their reference sheets throughout.

   Dr. Dreams, director of Catch Out

2. 2. One place per stretch of song

   Plate each stop empty before anyone stands in it, so the montage reads as a journey rather than a blur of backgrounds.

   Empty scene, no people or animals.

   Dr. Dreams, director of Catch Out

3. 3. Plan the change in the note

   Write a character's later look into the first note, so the second sheet is ready when the story reaches it.

   Long unkempt hair (later buzzed short).

   Dr. Dreams, director of Catch Out

4. 4. End wide and let the song finish

   Close on a wide shot with slow camera movement, so the last chorus or the credits have a picture to play over.

   Camera booms up slowly, rising above the train as it rolls off into the distance.

   Dr. Dreams, director of Catch Out

Questions about making an AI music video

More on making music videos in Flick

Film in this guide

• Dr. Dreams · Catch Out