Upload your song to the Atlabs Music Video workflow, pick Realistic and 16:9, write a one paragraph concept, and upload your character's portrait to the cast. Atlabs splits the track into scenes. Then fill each scene with a motion prompt, a library clip or lip sync, tagging each member by @name.
“Before the Glow” is a 66 second K-pop music video made in Atlabs, starring three punk idols. One of them is a custom character: her portrait was uploaded as a photo, so she appears in the video exactly as designed. The video starts from an uploaded song, “We Are Enough”. Atlabs cuts it into 23 short scenes, and the idols move from a black studio void through neon Tokyo streets, an underground garage and a concert stage before they unite in one formation for the chorus. This tutorial shows the setup, the cast and the locations, then every scene in order with how it was made, so you can build a personalised music video of your own.
What you'll need
You need a finished song, because its lyrics and beats become your scene list. You need a portrait of your own character, a photo of yourself or artwork you designed, plus portraits of any bandmates. And you need one sentence about what makes each member different. For this project that meant the track “We Are Enough”, a realistic look, and three idols with three energies: chaos, dominance and control.
Meet the idols
All three portraits were uploaded as images rather than written as descriptions, so no text prompt is saved for any of these characters. The picture itself is the reference Atlabs uses.
Minjun Choi (@MinjunChoi), the custom character
This is the personalised part of the video. Minjun's portrait was made outside Atlabs and uploaded as an image, and Atlabs uses it as her face and look in every scene. Her energy is dominance. She has long, full dark curls with a red glow on the edges, a red bindi, a nose ring and a septum ring, layered silver chains and a chest tattoo. She wears an oversized black jacket with red embroidery, studs and hanging chains over a black vinyl skirt, and stands with her arms crossed, chin up, looking straight at the camera.
Personalising a video is as simple as this. Upload a photo or artwork of yourself, or of a character you designed, as the cast portrait. Every scene then starts from that exact face.
Jihoon Park (@JihoonPark)
Jihoon's color is purple and her look is cyber idol. She has very long lavender hair with two space buns and glowing purple hair pins, dark purple lipstick and glossy lilac makeup. She wears a cropped purple moto jacket trimmed with blue LED strips over a mesh top and black vinyl shorts, with purple patent thigh high boots and LED platform sneakers.
Seungho Kim (@SeunghoKim)
Seungho's color is neon green and her look is street punk. She has a black blunt bob with neon green ends, green eyeliner, rhinestone face gems and dark plum lipstick. She wears a neon green trench coat over a fishnet top and a black vinyl skirt, with fishnet tights, black knee socks and combat boots with green straps.
The names in your cast don't have to match the look. Atlabs follows the portrait, so you can swap in a new image at any time and every new shot uses it.
The locations
No location was saved to this story's cast. The main setting comes from the concept text, “a black studio void under individual harsh spotlights”, and each scene's start frame adds its own place. The video moves through six looks.
The black studio void (Scenes 1 to 5 and 12 to 22): a glossy black floor, harsh spotlights and neon rim light. This is the home base of the video.
The underground garage (Scene 6): bare concrete pillars, strip lights and flying sparks.
The boiler room (Scene 7): dark machinery, steam and a cracked mirror glowing red.
The concert stage (Scene 8): one white spotlight from above and a crowd in the dark.
The neon Tokyo street (Scenes 9, 11 and 23): wet asphalt, rain and shop signs, straight from the lyric “Tokyo rain”.
The neon truss stage (Scene 17): the void again, now with purple and green neon tubes overhead for the chorus.
If you want to reuse a place across many scenes, save it as a location in the cast. You can tag it in prompts the same way as a character, with @Name.
Scene by scene: every lyric, frame and prompt
Scene 1 (2.0s): instrumental intro
Trimmed library clip, so no prompt is saved. The three idols stand in silhouette in the black void, lit only by the purple and green glow of their outfits, as the light slowly rises on them.
Open in near darkness. Holding back the faces for a few seconds makes the first close up hit harder.
Scene 2 (1.0s): instrumental
Library clip, so no prompt is saved. Jihoon kneels in her purple LED outfit, then the camera rushes in to an extreme close up of her face and violet eyes.
Scene 3 (1.0s): instrumental
Library clip, so no prompt is saved. The three idols stand shoulder to shoulder as streaks of white light behind them burst into purple, and their hair blows back.
Scene 4 (2.0s): instrumental
Library clip, so no prompt is saved. Minjun, in a black leather glove, rests her hand on her chin and stares into the lens under a red rim light.
Scene 5 (2.5s): instrumental
Library clip, so no prompt is saved. The three idols walk toward the camera in a V, with Minjun at the front.
Scenes 2, 4 and 5 introduce the group one look at a time: one member, then another, then all three together. It's a classic K-pop intro pattern.
Scene 6 (2.0s): “Three silver shadows step in Tokyo rain on our”
Library clip, so no prompt is saved. Seungho crouches in an underground concrete garage while sparks fly, and the camera rolls until the room is upside down.
Scene 7 (1.5s): “coats”
Library clip, so no prompt is saved. Jihoon leans on a cracked mirror glowing red in a dark boiler room, and the camera pulls back to show the whole room.
Scene 8 (3.5s): “So do Sao Paulo bling”
Library clip, so no prompt is saved. Seungho sings into a vintage microphone under a single spotlight, with a crowd in the dark behind her.
Scene 9 (1.0s): “Three hearts in a single…”
Library clip, so no prompt is saved. Jihoon snarls on a rainy neon street, her hair whipping sideways, as the camera pushes right into her face.
Scene 10 (1.0s): “…frame they look then”
Trimmed library clip, so no prompt is saved. Minjun grins wide with her hair flying in the red light.
Scene 11 (1.5s): “they lean back”
Library clip, so no prompt is saved. Seungho bares her teeth on a neon Tokyo street full of bar and ramen signs, and the camera pushes into a close up.
Scenes 9, 10 and 11 give each idol the same move, a snarl straight into a pushing camera, one second each. Repeating one action across the group is how you make three people feel like one band.
Scene 12 (3.5s): “Hands folded ice cold but we came with the same name”
This scene is a still image, not an animated clip. Its image prompt:
This is a good template for a group shot: lens, angle, setting, lighting, then who stands where by their @mention, then one shared expression and one shared move. The saved image shows three men in matching dark jackets rather than the three idols, so it looks like it was made before the uploaded portraits were added. If you copy this prompt, run it again after your cast is in place.
Scene 13 (4.0s): “We came to change the vibe”
Lip sync only, so no prompt is saved. Minjun sings the line straight into the camera under a red rim light.
Scene 14 (4.0s): “You said prove it so we did”
Lip sync only, so no prompt is saved. Jihoon sings to camera inside a red ring of light.
Scene 15 (3.5s): “You said break it why just leave”
Lip sync only, so no prompt is saved. Seungho sings to camera inside the same ring of light.
Scenes 13, 14 and 15 give each idol one line of the pre chorus, framed the same way. Lip sync each member on her own line and the group reads as a real vocal line up.
Scene 16 (3.5s): “One voice two voice three now the whole room starts to rise”
No prompt is saved. This image was uploaded and placed in the scene as a still: a three panel split with Minjun screaming, Jihoon staring into the lens and Seungho glaring, one panel per voice in the lyric. A split screen is the simplest way to show “one voice, two voice, three” in a single frame.
Scene 17 (4.0s): “We are enough”
The first chorus hit needs almost no prompt. The start frame already has the neon truss and the three idols walking forward, so the prompt just tells the camera to stay on their faces while they laugh and sing.
Scene 18 (3.5s): “We are enough”
A composite face, three faces layered into one, is the visual version of the concept line about separate identities dissolving into one force. Like Scene 12, the saved clip shows the earlier male version of the cast, so run it again with your own cast if you copy it.
Scene 19 (3.5s): “Say it back now”
Scene 20 (3.5s): “We are enough”
Small details like “a single bead of sweat on her temple” and “her chest rises and falls” make a still pose feel alive. This is Minjun's solo moment in the chorus.
Scene 21 (4.0s): “No we don't bend we don't hush”
“Syncs with the driving beat” tells the model the move should feel rhythmic. A circular orbit around a spinning group is the most dance like camera move you can ask for.
Scene 22 (5.0s): “We are enough”
End the chorus by fading the light, not cutting it. “The neon rim light begins to subtly fade” gives you a built in wind down.
Scene 23 (5.5s): outro
This scene is a still image, not an animated clip. Its image prompt:
The saved prompt has a gap before “frame center”, where a character @mention most likely sat and was removed. Put your own character's mention there, for example “@SeunghoKim frame center”, so the outro shows one of your idols. The saved image shows a different person, so make it again with your cast if you copy this scene.
What makes these prompts work
Start with a concept paragraph. One short paragraph about three energies and how they unite drives every scene, from the solo intros to the group chorus.
Personalise with an uploaded portrait. Upload your own character as a cast portrait and she appears as herself in every shot, with no description needed.
Color code the group. Purple, red and neon green make each idol easy to spot in one second cuts.
Write the camera like a cinematographer. Lens, angle, light and one move, like “35mm lens, low angle tracking forward” or “a rapid, tight circular orbit”.
Use Shot switch for two shots in one clip. Every chorus scene from 18 to 22 cuts from a wide or group shot to a close up inside one clip.
Lip sync each member on her own line. Scenes 13, 14 and 15 turn the pre chorus into a vocal line up.
Mix generation with editing. Library clips, trims, stills and a split screen all sit next to generated clips.
FAQ
Can I put myself or my own character in an AI music video?
Yes. Upload a photo or artwork as a cast portrait in Finalise Cast. Atlabs uses that picture as the character's face and look in every scene, as was done for Minjun Choi in “Before the Glow”.
How do I make a group look like one band?
Give each member one signature color, repeat the same move across all of them, and lip sync each member on her own line. “Before the Glow” does all three.
Do I need a prompt for every scene?
No. Most of the verse in “Before the Glow” uses library clips, three scenes are lip synced, and one scene is an uploaded split screen image.
What is Shot switch?
It puts two shots inside one clip. Write “Shot 1”, then “Shot switch”, then “Shot 2”, and the clip cuts between them, which is how the chorus gets its fast cuts.
Final verdict
“Before the Glow” shows how easy it is to star your own character in a K-pop style AI music video. Four setup steps in Atlabs did the groundwork: uploading the song in Add Music, choosing Realistic and 16:9 in Set Style, writing one concept paragraph in Creative Direction and uploading three portraits in Finalise Cast. From there, a three color code, library clips for the fast verse, lip sync for each member's line and Shot switch prompts for the chorus turned 23 scenes into one connected 66 second video. Upload your own portrait, use these prompts as templates, and tag each member by @name.
