A music video board fails when it looks right and moves wrong. The track has a structure (verse, chorus, break) and the board has to say what happens in each, or the beautiful references will produce a video that ignores the song. Build it against the timeline of the track, in three layers: performance, world, and rhythm.
Board against the track, not as a grid
Lay the board along the song's structure rather than as an undifferentiated wall. Verse, chorus, break, final chorus, each with its own small cluster of references and one line about what changes there.
That single structural choice does more than any reference selection. It forces the questions the video actually turns on: does the chorus open up or tighten? does the break go somewhere else entirely or stay and change the light? A grid of forty beautiful frames cannot answer either.
The three layers
Performance. Is the artist performing to camera, existing in a scene, absent entirely? This is the first decision and it constrains everything after. Show references of the treatment, not of the artist.
World. Location, palette, era, texture. The part everyone enjoys building and the part that means least on its own.
Rhythm. How the video cuts. Long takes, rapid montage, locked frames with movement inside them. Reference this with clips rather than stills where you can; rhythm is invisible in a still and it is half the video.
What to generate before there is a budget
Music video pitching is speculative by nature: directors present treatments for jobs they may not get, on budgets that may not exist. So the economics of proving the look matter more here than anywhere.
Canon prices on 8frame, a credit being $0.01 at pack rate: a still in the proposed treatment is 6 to 21 credits, and an 8-second moving beat on Veo 3.1 Lite is 54. A treatment with four style frames and two moving beats lands around 200 credits, roughly $2, which changes what an unpaid pitch can contain.
Use it for the world and the rhythm. Do not generate the artist: even for a pitch, a synthetic version of a real performer is a problem rather than a shortcut, and label anything generated clearly so the treatment is not mistaken for footage.
The board that gets a director hired
One image the whole video hangs on. Treatments that read as a collection lose to treatments that have a centre.
A rhythm reference that is actually moving. Two or three clips, not stills, showing how it cuts.
One idea nobody asked for. The reason to pick this director rather than the competent one.
Practicality named. How many locations, how many setups, what the schedule shape is. Treatments that ignore production read as fantasies, and labels have been burned by fantasies.
Where music video boards go wrong
Too many worlds. Three locations in a three-minute video is usually two too many, and boards happily promise five.
Referencing famous videos directly. Everyone in the room knows the reference and will compare the result to it forever.
Ignoring the lyrics. A board built purely on sonic feel can contradict what the song says. Read the lyrics once with the board open.
No plan for the chorus. The chorus repeats three times and cannot look identical each time. Board what changes.
FAQ
How long should a music video treatment be? Short. A page or two plus the board. Directors lose jobs to long treatments more often than to short ones.
Should I generate the artist in the board? No. Generate the world and show the artist through references of treatment. A synthetic version of a real performer creates problems that outlast the pitch.
How many references per section? Four to six per song section, cut hard. Forty references across the whole board is plenty and most treatments carry twice that.
Board the song's structure, prove the world cheaply, and keep the artist real. The 8frame canvas is free and unlimited, and generation is paid from $19/month.