How to make an AI music video from a photo and a song

01

Start with a passage you love

Upload a song you have permission to use and listen for a passage with a little shape of its own: a vocal entrance, a lift in energy, a line that lands. A short chorus is a good place to try your first scene. You can work with the whole song too.

For this example, we made a fifteen-second instrumental and used it from the beginning. There are no vocals to sync; the small lift in energy gives the performer something to move with. For your own song, choose a start and end, then listen once more. Leave the last word room to finish.

If there’s a moment you care about, tell Director: “Turn toward the camera when the melody comes in,” or “Hold the smile through the last note.” If your passage has vocals, add the lyrics and any timing notes you have, then check the proposed cues against the track.

Before you continue: The timings below describe our brief. Watch the result too: a requested gesture or cut may land differently in the generated clip.

02

Keep the details that make the performer yours

Choose a clear photo where you can see the face, hair and outfit. What would you notice if it changed? A silver jacket, round earrings, the shape of a haircut—pick a few details to carry through every shot.

Our practice performer wears a silver jacket and black trousers. Warm coral light spills out of the arcade; blue evening light catches the jacket outside. The lighting changes, but we should still recognize the same person. Check that first look before making the rest of the scene.

• Learn how to review character continuity →

03

Give the performance one small story

Picture the arcade door opening. The performer steps into the street, notices us and turns toward the lens. A relaxed shoulder movement and hand gesture follow the music. Then they spot a friend just beyond the camera. A small smile gives us an ending. We can do all of that in one place.

The door gives us an entrance. The performance brings us closer. The friend gives the final look a reason. These little actions are easier to direct than a long list of moods and camera effects.

• Turn your visual beats into a storyboard →

04

Find the energy in small details

If the song wants a faster pace, look around the scene: shoes splashing through a reflection, a hand catching the light, a friend laughing. A brief cutaway can lift the energy while the vocal keeps playing. Come back to the performer so we know whose night this is.

Cuts every 0.5–2 seconds are choices you make in the edit. Start with usable passages of movement, then pick the moments worth keeping. You don’t need to ask for a separate half-second generation for every cut. Let the final smile stay on screen long enough to feel it.

05

Tell Director what you see

You have a passage, a person and a small story. Put them together in a few sentences. Here’s the arcade idea ready to adapt; change the clothing, place and times to match your files.

06

Try the performance shot, then watch the whole scene

Make the middle shot first: it asks the most of the performer. Watch at normal speed with the song playing. Does the hand gesture feel natural? Is that still the same face and jacket? If your version includes singing, check the mouth against the vocal too. Revise anything distracting before carrying the approach into a longer video.

In our result, the silver jacket and street stay recognizable, and the ending settles on a smile. The performance is restrained; for a more energetic song, we’d ask for a clearer movement and test that shot again.

Once you have the scene, watch the cuts. The arcade door and street should feel connected, and the smile should arrive with enough time to settle. If the pictures finish early, ask for a visual adjustment that keeps the song’s chosen length and original speed.

Then listen without looking. You want one continuous recording, with no doubled vocals or repeated words. If you want a song-only mix, the request below tells Director exactly what to keep. Play the finished file from beginning to end before sharing it.

• Revise one shot while keeping the rest of your film →