How to Make a Music Video That Actually Looks Sick

Most weak AI music videos fail before the model renders anything. The prompt describes a mood, but it never directs a shot.

Dark, cinematic, emotional is not a visual plan. A useful brief names the world, the subject, the light, the camera, and the motion.

Start with one visual world

Write one sentence that could guide the whole release:

A lone dancer crossing an empty blue subway platform at 2 a.m., hard fluorescent light, wet concrete, slow handheld camera.

That sentence has five things the model can use:

• Subject: one dancer
• Place: an empty subway platform
• Time and light: 2 a.m. under hard fluorescent light
• Texture: wet concrete
• Camera behavior: slow handheld movement

Keep that world consistent across the cover, profile image, Canvas, and music video. A release feels expensive when every asset appears to come from the same place.

Direct three shots

You do not need fifteen unrelated generations. Start with three shots that have different jobs.

1. Establish the world

Show where we are. Use a wide frame, a clear silhouette, and one strong source of light.

Wide shot of a lone dancer on an empty subway platform, deep cobalt shadows, fluorescent ceiling lights receding into the distance, slow push forward.

2. Give us one detail

Move close enough to make the scene physical.

Close shot of scuffed sneakers sliding across wet tile, reflected blue light, shallow depth of field, camera tracking beside the movement.

3. Land the payoff

Save the biggest change for the hook, drop, or final beat.

The train tears through the station behind the dancer as every overhead light flashes on in sequence, locked camera, wind moving their coat.

Use motion verbs, not mood words

Motion prompts get clearer when the verbs describe something a camera or subject can actually do.

Use words like:

• tracks
• pushes in
• circles
• drifts
• snaps into focus
• crosses frame
• flickers
• ripples
• falls away

Avoid filling the prompt with epic, stunning, vibey, premium, or cinematic. One camera direction is worth more than five adjectives.

Pick the format before the shot

Music House supports vertical, square, and wide music-video formats.

Format	Use it for	Compose around
9:16	Reels, TikTok, Shorts	One central subject and safe space near the edges
1:1	Feed posts and release tiles	A strong center with room to crop
16:9	YouTube and wide players	Environment, movement, and lateral composition

Do not generate a beautiful wide shot and hope a vertical crop will rescue it. Choose the frame first.

Keep text out of the generated footage

If the video needs a title, date, or artist name, add it after the visual is rendered. Generated typography can drift between frames. Clean footage also gives you more ways to reuse the same shot.

A prompt shape that works

Use this order:

Shot size + subject + specific place + lighting and palette + physical action + camera movement

Example:

Medium profile shot of a singer alone inside a closed laundromat, midnight cobalt and warm machine light, condensation on the windows, she turns as every dryer stops at once, camera slowly circling clockwise.

Open Music Video Maker, add your audio, choose the format, and direct the first shot before you build the rest.