Cuecraft: the AI Director for live events

Cuecraft's dashboard: a still frame from the crowd camera, the seven analysis stages all complete, a table of 773 decisions with the reason and confidence for each, and the whole night's cut timeline across the bottom.

The actual app, on an actual 77‑minute night. Every row in that table is a cut it made, and the paragraph under the frame is the arithmetic it used to make it.

This is what Cuecraft actually does

773 cuts. Three cameras, six files. One night, and nobody touched a keyboard.

dj · 319 crowd · 299 roam · 155 773 shots · 77 minutes · mean shot 6.0s · opacity is each cut's own confidence

One whole night, 77 minutes, three cameras across six files, cut automatically into 773 shots. The band under the timeline is the story arc the app read off the music by itself: release, build, peak.

Now closer. Ten minutes of the same set.

dj · 28 crowd · 15 roam · 12 55 shots · 112 BPM · opacity is each cut's own confidence

The same night, ten minutes of it up close, over the master audio it was cut to. Hover any shot to see which camera it chose and how sure it was.

Every block is one shot. As wide as the shot is long, coloured by the camera it chose, over the master audio it cut to. Hover any of them to see what it picked and how sure it was.

Understand the moment. Choose the shot. Build the edit.

Cuecraft does not switch cameras on a timer. It reads what the performance is doing and makes a directing decision about it.

How it works

Three things happen while you make coffee.

01

It understands the performance

Every camera is lined up against the master recording by its own audio, and the alignment is checked rather than trusted. Then it watches every second of every angle, scoring sharpness, exposure and steadiness, while reading the music underneath for the builds, the drops and the moments the room reacts.

The detail

Each camera is judged against how it normally looks, never against the camera next to it, so the soft lens in the corner still earns its cuts on the shots where it is genuinely the best thing in the room. Sync is verified end-of-set as well as at the start, which has caught drift on footage that was described as already synced.

02

It chooses the shot

Cut points come out of the music itself, not a metronome. At each one it weighs every camera and picks, instead of you scrubbing ninety minutes three angles at a time deciding the same thing by hand.

The detail

Every cut carries the numbers it was chosen on, in plain language. When you disagree with one you can read exactly what it saw, which means you know whether to trust the next one.

03

It hands you an editable Resolve timeline

Not a rendered video. Not a locked result. An FCPXML that imports as a flat, normal, editable track with a colour group per camera role, where every cut is a cut you can drag.

The detail

Nothing is transcoded, nothing is re-encoded, and no pixel of your footage is touched. Cuecraft never renders video, which is also why there is nothing to wait for at the end.

Event awareness

One moment. Three cameras. One decision.

Here is a real cut from the timeline above, at 17.9 seconds, with the actual numbers the app recorded. The highest-scoring camera did not win.

17.9saudio onset

dj0.78 on air 17.9s

crowd0.45

roam0.72 selected

Real values from decision_list.json, shot 2 of 55.

DJ scored highest. Roam got the cut, because the DJ camera had been on air for eighteen seconds and the music had just given it a reason to move. That is the difference between knowing a cut should happen and deciding what the audience should see.

And it was close, so it says so: confidence 0.41, won by 0.041. A cut it is unsure about is one you might want to look at, and it tells you which ones those are.

Director DNA™

Every editor sees the same night differently.

So Cuecraft does not have one correct way to cut a set. Every decision it makes comes out of a named set of numbers called your Director DNA, and those numbers are yours to move.

default-v0

min shot          4.0s      how short a shot may ever be
cooldown          2.0s      quiet time after a cut
max shot         20.0s      nobody holds longer than this
variety push      5.0s      how fast it gets restless

phase pacing      peak 1.4  build 1.0  release 0.7
camera bias       dj 0.00  crowd 0.00  roam +0.05

The phase line is the event-aware part. Cuecraft knows the difference between a build, a peak and a release. Those three numbers are you telling it what to do about that: cut 40% harder through a peak, settle right down through a breakdown. Move one, re-run, and the whole set re-cuts to match.

Every export records which DNA produced it, so two versions of the same night are never a mystery.

Director DNA ships today, and you tune it yourself. The step after this one is Cuecraft learning from the cuts you keep and the cuts you throw away, so you stop touching the numbers at all. That part is being built and is not in the app yet, and we would rather tell you that than let you find out.

Proof

Not a demo. Not a mockup.

The timeline at the top of this page is a real export from a real club night, produced by the shipping app and pasted here unedited.

10minutes

3cameras

55shots

112BPM read

0manual cuts

Why did it choose this shot?

This is the app's own record for one cut, in its own words:

Cut on an audio onset at 82.9s, 16.6s after the last one
Chose crowd, scoring 0.71
Sharpness 0.96, exposure 0.59, steadiness 0.58
Next best was roam at 0.58
Confidence 1.00, won by a margin of 0.135

Show the raw reasoning

cut on audio onset at 82.9s (16.6s since last cut,
  min shot 4.0s, cooldown 2.0s)

crowd 0.71 = quality 0.66
  [blur 0.96*0.20 + exposure 0.59*0.40 + shake 0.58*0.40]
  + prior +0.05

confidence 1.00 - won by 0.135 over the next best
  option, runner up roam (0.58)

That is the whole difference between this and a magic button. You are not being asked to trust a model that cannot tell you anything. You can audit any cut, decide it was wrong, and change it.

The honest bit about proof: this is new, so there is no wall of testimonials and none are going to be invented. What there is instead is the free tier. Do not believe me. Run it on your own footage.

Local and private

Your footage stays on your machine.

Cuecraft runs entirely on your Mac. There is no upload path in the software, no cloud processing and no requirement to send a client's night to a third party server. The only thing that ever leaves your machine is the email address you buy with.

Local processingNo uploadWorks offlineNo account

Stated precisely because it is checkable: the application makes no network requests at all. It works with the venue wifi down and the laptop in flight mode.

Try it

Try it on your own footage. No card.

The free version is not a crippled demo, and it is not a fifteen minute tool. It ingests your whole set, syncs every camera and cuts the entire timeline, however long the night is. A three hour set is cut in full, same analysis and same quality as paid. What is gated is only the export: the first fifteen minutes of that timeline open in Resolve, which is enough to judge whether the machine cuts the way you would. Paying releases the rest.

	Free	Paid
Sync, analysis and cutting	Everything	Everything
Set length	Any. Three hours, or more	Any. Three hours, or more
Cameras	Three: dj, crowd, roam	Three: dj, crowd, roam
Resolve export	First ~15 minutes	The whole set
Colour group per camera role	Yes	Yes
Runs offline, footage stays local	Yes	Yes
Card required	No	Yes
Price	Free, permanently	£25/mo or £250/yr

If Cuecraft does not make your edit faster, do not pay for it. Fifteen minutes is enough to see whether the machine cuts the way you would, on your own footage, before anybody has your card details.

Pricing

One product. Three ways to have it.

Free

£0 / forever

Your whole set cut, however long it is. The first fifteen minutes export to Resolve. No card, no expiry.

Get early access

Monthly

£25 / month

Cancel any time. If you only shoot in season, pay in season.

Start free trial

Two months free

Yearly

£250 / year

The same thing at £20.83 a month.

Start free trial

There is no hour counter. Cuecraft runs on your own machine, so an hour of footage costs nothing to process and nothing to meter. The cloud tools price by the hour — twenty dollars a month typically buys ten hours of media, and one three-camera night is most of that. Here a night is a night, and there is no month where you have run out.

Both paid plans start with 14 days free, and a card goes on at that point so the trial rolls into the subscription without you thinking about it again. You will get an email before the first charge, not after. Cancel inside the fourteen days and you are never charged. The free tier needs no card at all and never expires.

If it does not do what this page says it does, email within 30 days of a charge and it is refunded. When a subscription ends the app drops back to free and keeps working, and everything you already exported is yours and unchanged.

Straight answers

The things you are already wondering.

What happens if Cuecraft makes a bad cut?

You change it. It is an editable Resolve timeline, not a rendered video, and that is deliberate. Worst case you disagree with some shots and move them, on a timeline that is already synced, organised and roughly cut. You are correcting a first pass instead of building one.

What cameras does it support?

Three, in fixed roles: a DJ camera, a crowd camera, and a roaming camera. Any camera Resolve can read works, and they can be different bodies at different frame rates and resolutions, because that is what real shoots look like. A fourth role is a product decision rather than a setting, and it is not in the app today.

Does it work with footage that is already synced?

Yes, and it checks anyway. Sync is verified rather than trusted, at the end of the set as well as the start. That check has already caught drift on footage that everyone involved believed was fine.

How long does a one-hour event take to analyse?

Roughly half the length of the set on the machine it was built on, and most of that is generating proxies the first time it sees the footage. Proxies are cached, so a re-run with different settings is minutes rather than tens of minutes. That is one measurement on one Mac, not a promise about yours.

Does my footage go anywhere?

No. There is no upload path in the software at all, and it makes no network requests.

Do I need the paid version of Resolve?

No. It exports a standard FCPXML and uses colour grouping, both of which the free Resolve supports.

Is this going to replace me?

It replaces the part of your job that is clicking, not the part that is taste. It does not grade, does not pick your look, does not do your sound, and will not rescue footage that was out of focus on the night.

Made by editors, for editors

I built this because I could not face my own footage.

The shoot is the easy bit. You get the angles, the room goes off, and you go home at three in the morning with cards full of footage you are genuinely excited about.

Then the cards go on the desk and sit there.

The grade is creative. The pacing is creative. The look is yours. But that first mechanical pass, the one where you just decide which camera is on screen, is not. It is a couple of days of clicking with a soundtrack.

I got tired of it, and I am a programmer as well as an editor, so I built the thing I wanted to exist. It has been cutting my own sets for months. That is the entire credential: I made it for me, it works on my footage, and I am now letting other people have it.

The longer version

Because you know what is coming. Pulling three cameras in. Lining them up against the master audio and finding out that two of them drifted. Building a multicam clip. Then scrubbing through ninety minutes of it, three angles at a time, deciding over and over which one is worth looking at, while the same track loops in your head until you hate it.

So the job you quoted as a one night shoot becomes a fortnight of evenings. The client starts asking. You start apologising. And the next enquiry comes in and some part of you hopes it is a wedding instead, because at least a wedding has one camera pointed at the thing that matters.

That is the part nobody puts on a pricing page: it is two weekends per set, every set, for as long as you keep shooting them. It is the enquiries you turn down because you are still buried in the last one. And it is what happens to editors who spend years doing the mechanical pass by hand, which is that eventually they stop taking the jobs they used to love, because the shoot is four hours and the edit is forever.

Your next set is already waiting.

Don't spend another weekend deciding which camera should be on screen.

Cuecraft ships soon, free tier included. One email the day it does, nothing before, unsubscribe in one click.

P.S. If you only read this far: the timeline at the top of this page is a real ten minutes of a real club set, cut automatically, and you can hover any shot to see what the machine chose and why. When it ships there is a permanently free version, no card, that cuts your whole set and exports the first fifteen minutes, so you can check it against your own footage without paying anybody. Put your email in the box above.