a change is coming to luttie

bringing the news about what i'm working on before it drops.

i'm rebuilding the way luttie grades an image, and i'd rather tell you while it's still being built than present it finished.

what luttie is right now

luttie currently has a linear color grading process, putting images through a fixed sequence of adjustments. white balance, then exposure, then contrast, then the colour wheels, then blacks and whites, saturation, vibrance, your lut, the hsl secondaries, curves, vignette, and finally the film effects - bloom, halation, grain.

every lut anyone has ever exported from luttie went through those stages in exactly that order or some variation but once settings are applied it's difficult to modify them in a satisfying way.

for the majority of luttie's users, that's fine. you want a look, you push some buttons/sliders, you export a .cube and get on with your edit. a fixed chain is faster for that, not slower.

it has obvious limits, and that's what i'm trying to improve.

you can't warm shadows without also touching the skin tones that live in the same tonal range. you can't take one adjustment, mask it to a corner of the frame, and leave the rest alone. you can't build a look in two halves and blend between them. this is possible in a real grading suite, and none of them are possible using luttie's current architecture.

what's changing

i'm rebuilding the engine to allow for more freedom with the grading.

the plan is to break that fixed sequence into individual operations you can reorder, disable, mask, and eventually branch, structure more like a graph than a line. if you've used davinci resolve's node editor, then you'll know what i mean. it'll be a challenge to get this working in the browser, but if it can be done, it could be huge.

three things have to change under the hood for that to work, and one of them is worth mentioning even if you never touch a node:

the image will be processed in higher precision. right now every stage hands off to the next in 8 bits per channel. that's fine for a short chain, but for complex jobs where you're stacking dozens of operations, each one rounding its result, you start seeing banding in skies and gradients. the new engine keeps everything in 16-bit float until the very end. that's a straightforward quality improvement to every grade luttie produces, whether or not you ever rearrange anything.

some operations will stop fitting in a lut. a .cube file maps a colour to a colour — it has no idea where a pixel sits in the frame. so anything spatial (masks, shaped windows, real blur-based halation) genuinely cannot be baked into a lut export, and i'd rather tell you that plainly in the ui than quietly export something that doesn't match what you saw. lut export isn't going anywhere; some of the new tools just won't be part of it, and luttie will say so. what i'm building is an extension to what exists already so you can still create luts the way you did before.

it'll be built in stages. roughly in this order:

1. a reorderable grade stack — every adjustment becomes its own step you can drag into any order, or switch off without losing its settings.
2. masks and windows — limit any single step to part of the frame, so you can warm a background without warming a face.
3. branching and blending — build a look in two parallel halves and mix between them, instead of one operation always feeding the next.

no dates. i'd rather ship the first one properly than promise all three by a month i'll miss.

what this doesn't change

the editor you use today stays exactly as it is.

i want to be clear about this, because "we rebuilt the engine" usually means "your tool now looks different and you have to relearn it." that's not what's happening. the sliders, the wheels, the curves, the preset browser — same place, same behaviour. if you open luttie the week after this ships and never touch anything new, you shouldn't be able to tell it changed.

the new tools will most likely live behind an advanced mode toggle. off by default. flip it on when you want the extra control, flip it off when you just want a look and a .cube file.

why i'm telling you before it's done

two reasons.

the first is that i'd rather build this in the open than present it finished. luttie is small enough that you can see my hands on it, and i think that's a feature, not something to hide behind a polished changelog.

the second is accountability. it is much harder to quietly abandon something you've written down in public.

so — if there's something you've tried to do in luttie and couldn't, i want to hear it. that's the most useful thing you could send me right now, because it's the difference between me building the right three features and the wrong ones. my dms on x are always open.

- dom