When Your ACES IDT Mismatch Creates a Color Shift That Follows You to the DI

You're three weeks into the grade. The director loves the look. Then the DI session starts, and suddenly skin tones shift magenta on the big screen. The producer panics. The colorist swears nothing changed. But something did—and it's been hiding in your ACES pipeline since the first IDT assignment.

This isn't a bug report. It's the logical outcome of a mismatch between the Input Device Transform written into your clip metadata and the actual sensor response of the camera that shot it. When that gap exists, every grade you apply is built on a false foundation. And because ACES is designed to be linear and predictable, the error doesn't get buried—it gets amplified, carried through the entire pipeline, and eventually lands on the DI screen. The fix isn't complicated. But finding it before the final render? That takes knowing where to look.

Who has to choose—and by when

The Colorist’s deadline: pre-grade vs. post-edit

The decision lands on the colorist’s desk before the first node opens. Not during the creative pass, not when the client is watching — before. If the IDT doesn’t match the camera that shot the footage, every curve, every LUT, every Power Window you drop is built on a lie. I’ve watched a senior colorist spend four hours fixing skin tones that were cyan-shifted from the start, and the fix was a single dropdown menu that should have been set on ingest. The deadline? Before you grade. Once you commit to a grade, the mismatch embeds. It becomes part of the look. Reassigning later means re-grading every shot — or living with the error. Most facilities don’t have the budget for that. So the window is tight: between the conform and the first color pass. That’s when the decision gets made, and if it’s wrong, the DI suite becomes a detective agency, not a creative studio.

The DIT’s window for metadata checks

On set, the DIT has the cleanest chance to catch an IDT mismatch. They’re the ones who see the raw files first, the ones who can crack open the metadata and confirm the camera model, the log curve, the color space. That window closes fast — usually before the card leaves the cart. If the DIT flags a mismatch and reassigns the IDT in the dailies pipeline, the editor never sees the shift. The colorist never fights it. But here’s the catch: many DITs are told not to touch metadata. ‘Don’t change anything, the colorist will handle it’ — I hear that line every month. It’s a risk. A metadata check takes thirty seconds. A full DI reconform with baked-in mismatches can eat a day. The DIT’s window is the only one where the fix costs nothing. After that, you’re paying in time or quality.

Producer’s tolerance for re-conforms

Producers hate re-conforms. Hate them. Not because they’re technical — they don’t care about IDTs — but because re-conforms cost money and kill schedule. If the IDT mismatch is caught late, say during the DI, the producer has to decide: live with the shift or pay for a fix. That’s a brutal conversation. I’ve seen producers wave off a slight green cast because the alternative meant recutting three scenes with VFX plates that were rendered with the wrong IDT. The tolerance is low — zero, really — for anything that touches the locked edit. So the hidden deadline is the lock. Before picture lock, reassigning an IDT is just an EDL change. After lock, it’s a negotiation. The producer’s tolerance runs out the moment the conform is signed off. That’s when the colorist inherits a problem that should have been solved by the DIT, or worse, by the camera rental house who sent the wrong look file.

The fix is simple on paper: match the IDT to the camera. But who owns that choice? The DIT? The colorist? The producer? The answer is yes — and the deadline hits each of them at a different point. Miss yours, and the shift follows you into the DI. Not yet. Wait until you read the fix options.

Three ways to fix an IDT mismatch

Manual IDT reassign in the timeline

The fastest fix — and the one most colorists reach for first — is to manually swap the IDT tag inside the grading software. Resolve, Baselight, and Flame all let you right-click a clip, open the color-space metadata, and pick a different camera input transform. Wrong IDT baked in? Change it. The software re-runs the math as if the clip had been tagged correctly from the start. I have fixed entire seasons this way: a documentary where the DP shot Arri Alexa but a DIT accidentally tagged everything as Sony Venice. That shift? Instant coral skin tones and clipped highlights. A quick manual reassign on the master timeline, and the footage snapped back.

The catch is precision. Manual reassign works only if the original raw or Log file carries enough headroom. If the footage was already clipped or had a LUT baked in downstream, you can't un-ring that bell. And you need to know — really know — which camera actually captured each clip. Wrong IDT reassign creates a second mismatch on top of the first. That hurts. One post house I worked with guessed the camera model from the file name and ended up with a green/magenta cross that took hours to chase.

Per-clip correction with ACESclip metadata

ACESclip is the unsung hero here — an open-standard sidecar file that holds per-shot color metadata without touching the pixel data. Instead of reassigning the IDT globally, you write an ACESclip that overrides the input transform for specific clips. The grading app reads it at conform time and re-interprets the source. Why bother? Because you can fix one clip in a timeline of 800 without breaking the others. That's invaluable on multi-camera shows where a single camera had a wrong firmware look baked in.

Most teams skip this: they don't know that ACESclip can hold custom CTL transforms, not just presets. So you could point it to a corrected IDT written as a CTL file and have the system apply it frame-accurately. The trade-off is workflow overhead. You need to generate those sidecars, name them correctly, and get them into the conform pipeline before color begins. Miss one, and that clip stays broken. I have seen a feature where the ACESclip folder had a single typo in the reel name, and six shots landed on the wrong IDT. The fix took twenty minutes to find but three seconds to apply once spotted.

Not every film checklist earns its ink.

Not every film checklist earns its ink.

Custom IDT rebuild from camera measurements

When neither reassign nor ACESclip works — when the camera is off-menu, or the manufacturer's IDT is simply wrong for that specific unit — you rebuild the IDT from scratch. This means shooting a color chart under controlled lighting on that exact camera body, extracting the raw sensor responses, and generating a new input transform using CTL or a color-science tool like Colour Science for Python. Is it overkill for a single music video? Yes. For a branded content series that must match across five different camera models over three months? It's the only sane path.

'We rebuilt the IDT for the Sony FX6 on a car shoot. The factory IDT was flattening the red channel in tungsten. Once we replaced it, the skin tones stopped looking like raw chicken.'

— Technical colorist, automotive commercial, 2024

The pitfall is time and expertise. You need the chart shots, the raw files, and someone who understands matrix math well enough to write a transform that doesn't introduce new errors. Most colorists never do this. But when you do — when you hand a colorist a custom IDT that matches the on-set reality — the DI session becomes about art, not about chasing a color shift that should never have existed.

How to compare the options

Turnaround time vs. accuracy trade-off

The fastest fix isn't always the right one — and that tension is where most teams get tripped up. If you reassign the IDT manually in your color-managed timeline, you can clear a mismatched clip in under thirty seconds. Click. Done. That speed matters when you're staring at a 400-clip timeline and the colorist is waiting. But here's the rub: manual reassign only works if you know exactly which camera original IDT belongs to that footage. Guess wrong — or grab a generic "Alexa Wide Gamut" that's close but not identical — and you're merely swapping one color shift for another. The alternative, rebuilding the IDT from a known target, takes hours or days. You shoot a ColorChecker, export a frame, run it through a transform builder, test the result against grayscale patches. Tedious. Yet that method catches weird edge cases — like a camera that was actually recording in S-Log2 but tagged as S-Log3 in the metadata. I've seen a single mislabeled clip poison an entire DI grade because nobody stopped to verify. So ask yourself: do I need "fast enough" or "provably correct"?

Skill level required for each method

Manual reassign is dead simple — anyone who can open a project settings panel can do it. That's its appeal. But the catch is hiding in that "anyone" part: the person clicking the dropdown needs to understand color science basics. What's a gamut boundary? Why does a LogC-to-Linear mapping break when fed ARRIRAW data? Most assistants I've worked with can follow a checklist, but fewer can spot why a skin tone shifted magenta after reassignment. That's the pitfall — you fix the metadata mismatch but introduce a transform mismatch. Rebuilding the IDT, meanwhile, demands a solid grasp of ACES, exposure math, and often a bit of Python or CTL scripting. Not every post house has that skill in-house. If you're a solo colorist on a tight budget, manual reassign is your only realistic option. But if you have a pipeline T.D. or a dailies color scientist on retainer, the rebuild route becomes viable — and safer. The wrong choice here can cost you a full day of re-conforming later.

Number of clips and scene complexity

One mismatched shot in a thirty-second commercial? Manual reassign, no contest. You spend more time arguing about whether to fix it than actually fixing it. But scale changes everything. A 200-clip documentary shot across five different camera brands, each with nested re-encodes and mixed frame rates — that's where complexity compounds. The scene complexity matters too: a talking head against a gray wall hides IDT errors better than a saturated product shot with brand-color-critical reds. I once had to bail out a feature where the drone footage IDT was wrong across thirty-seven aerial plates. Each clip looked fine in isolation. Cut together, the sky shifted from teal to cyan every time the drone shot appeared. We couldn't reassign each one manually — too many edge cases with highlight roll-off. So we rebuilt a single custom IDT for the entire drone package. Took two days. But the DI went clean after that. The rule of thumb: if the mismatch touches more than ten clips or any clip with hard-to-match colors (neon signage, green-screen backings, skintones in mixed light), lean toward rebuild. Manual reassign is a band-aid. Rebuild is surgery — and you don't want surgery for a paper cut.

You can't fix a pipeline problem with a dropdown menu, but you can certainly make it worse.

— senior dailies colorist, on why she insists on verifying IDT metadata before ingest

Trade-offs at a glance

Speed vs. precision in manual reassign

Manual reassign looks tempting on paper—a single dropdown click, and you're done. That's the speed. I've seen a colorist swap an Alexa IDT for a Sony Venice in under thirty seconds during a live session, grinned, and called it fixed. The catch is brutal: you're guessing which transform the original footage actually needed. Most DITs tag clips with a camera model, but not always the exact color science version. Pick the wrong ACES Input Device Transform and you've just swapped one color shift for another. What usually breaks first is the skin tones—they go waxy or cyan-tinted in the shadows. Manual reassign works when you have pristine metadata and a known camera path. Otherwise, you lose a day matching shots that should have matched out of the box.

That sounds fine until you're juggling three camera types on one timeline. The precision gap shows up in the grade: highlights clip differently, black levels drift, and the seam between two supposedly identical cameras blows out. Manual reassign doesn't fix the underlying transform error; it only changes which transform you're betting on.

Metadata dependency in ACESclip

ACESclip promises better accuracy by embedding the IDT choice directly into the clip metadata. The pros are real: once assigned, the transform travels with the file from set to DI. No guessing, no re-picking. But metadata dependencies have a dark side. If the ACESclip metadata gets stripped during a transcoding step—and I have watched exactly this happen on a TVC job—you're back to square one. The clip loads with a default transform that doesn't match, and nobody notices until the first client review. The tricky bit is that ACESclip assumes everyone downstream respects the metadata. Most pipelines do, but one QuickTime export with "flatten metadata" ticked? That hurts. The trade-off is reliability inside a controlled workflow versus fragility when files leave that bubble.

Reality check: name the production owner or stop.

Reality check: name the production owner or stop.

'Metadata is a promise, not a guarantee. One rewrap can orphan your entire transform chain.'

— colorist, post on LiftGammaGain forum, 2023

Cost and complexity of custom IDT

Building a custom IDT is the nuclear option. You send a camera to a lab, shoot a color chart under controlled lighting, and get back a precise mathematical inversion of the sensor's response. No mismatch. No guesswork. The cost, however, hits three places: the lab fee (easily $1,500–$4,000 per camera), the time (two to three weeks if the lab is busy), and the complexity of maintaining that IDT across firmware updates. Most teams skip this because they're on a tight turnaround. But here is where I push back: if your show has a single primary camera that will generate 80% of the final runtime, a custom IDT pays for itself in saved conform time. The real trade-off is not just money—it's the risk of locking into a transform that becomes obsolete when the camera manufacturer updates its color science. Are you ready to rebuild that IDT every six months?

What to do after you choose

Applying the fix in Resolve

Once you've settled on manual reassign or a full rebuild, the clock is ticking. I've watched teams spend forty minutes debating which route—then freeze when they actually have to execute. Don't be that room. Open Resolve's Color Management panel first. If you're reassigning, right-click the source clip in the Media Pool, select 'Input Color Space,' and override the metadata. That's it—two clicks, provided your timeline is set to ACEScc. The odd part is: most people forget to check 'Use Separate Color Space for Input.' Without that toggle, Resolve ignores your manual override. You'll think you fixed it, but the IDT mismatch still bleeds through. Verify by loading a grayscale ramp—if the neutral axis shifts green or magenta, your override isn't sticking.

For a rebuild, you're essentially creating a new IDT from scratch. Export a 16-bit DPX or EXR frame—ten seconds of a color chart plus skin tones—and feed it into ACES's CTL-based transform builder. Painful? Yes. But when the original camera log is broken or undocumented, it's the only path that doesn't gamble. The rebuild spits out a .ctl file; drop it into /Library/Application Support/ACES/transforms/ (or your OS equivalent). Restart Resolve, assign the custom IDT, and pray nothing chokes on the LUT precision. That sounds grim—it usually works, but the first pass often needs a tweak on the toe contrast.

Verifying with a known color chart

Here's where most teams skip a step. They apply the fix, glance at the waveform, and call it good. Wrong order. You need a locked-down reference: a Macbeth or DSC Labs chart shot under the same lighting as the problematic footage. Load it into a fresh timeline, same ACES settings, and read the RGB values off the neutral patches. If your fix is correct, the grayscale squares should fall within ±0.5% of each other—any wider and you're hiding the mismatch, not solving it. The catch is that skin tones can fool you; they mask a mild cross-talk that blows open in saturated reds or blues. I've seen a grade pass QC only to fail later because a crimson jacket turned brick-red. Trust the chart, not the face.

‘The chart never lies. The monitor's gamma setting, your viewing environment, and your fatigue level—those lie constantly.’

— Senior colorist, post-production facility in London

Run a second check: export a still, load it into a color analyzer like ColourSpace or LightSpace, and plot the delta-E against the chart's known values. If any patch exceeds ΔE 2.0, your IDT fix is still leaking. Don't chase the grade yet—go back and confirm the transform's matrix coefficients. The most common pitfall? Setting the wrong white point. ACES assumes D60, but your camera's metadata might report D65. That three-percent shift turns into a visible blue cast on the final print.

Locking the grade and handing off to DI

Once the numbers line up, lock the grade. Export a clean EDL with the IDT override baked into the clip metadata—don't rely on the DI team having your exact Resolve database. Send a single-frame reference EXR with the color chart visible in the corner as a sanity check. I always include a short readme: 'IDT reassigned to ARRI Alexa LogC; timeline is ACEScc; don't auto-detect input.' That document saves hours of back-and-forth. The real risk? The DI house re-ingests your files and the system auto-assigns the original, broken IDT. You show up to the final screening, and suddenly the black levels crawl into blue. That hurts. So specify in the delivery template: 'Input color space = forced override. No metadata trust.' Then walk away—you've done the fix. Let the colorist's work speak.

Risks of ignoring the mismatch

Wasted grading time on false foundation

You spend hours dialing in a look—skin tones, contrast, that signature teal-orange push—only to realize the IDT was feeding your grade through a distorted lens the whole time. The catch is brutal: every curve you pulled, every secondary correction you layered, was compensating for a shift that shouldn't exist. I've watched colorists burn three full sessions on a hero shot, then reassign the correct IDT and watch the whole grade collapse. That's not lost effort—it's lost trust in your own eyes. The foundation was wrong from ingest, and no amount of artful grading can fix a broken math pipeline. What usually breaks first is the black point: it drifts, you pull it back, then the midtones go muddy. You chase your tail until the producer asks why the last five versions all look different.

Client friction during DI review

Nothing sours a DI session like the director saying "the sky looked right on the monitor yesterday" and you have no good answer. The mismatch creates a color shift that follows you to the suite—but inconsistently. It might look acceptable on your calibrated OLED, then bloom green on the theater projector. The client sees a problem; you see a ghost you can't exorcise without admitting the IDT was wrong from the start. One production I consulted for shipped a trailer with a 2-stop blue cast in the shadows—nobody caught it until the theatrical QC report. — freelance color scientist, theatrical QC report

Odd bit about production: the dull step fails first.

Odd bit about production: the dull step fails first.

That friction escalates fast. You can't blame the monitor, you can't blame the LUT, and you definitely can't tell the director "the math was wrong in a dropdown menu six weeks ago." So you compensate—which leads us to the next trap.

Band-from over-correction when you compensate wrong

Most teams skip the root fix and just push a global tint adjustment in the grade. That sounds efficient until you hit a dissolve or a backlit close-up—then the banding appears. The over-correction amplifies quantization errors because you're stretching 10-bit data through a 14-bit pipeline built on bad math. The result? Visible contouring in gradients, false colors in shadow roll-offs, and a shelf of grain that looks unnaturally crunchy. I fixed one of these by going back to the raw IDT reassignment, and the "noise problem" that had eaten three weeks of denoising budget simply vanished. The pitfall is seductive: it's faster to grade around the problem than to re-ingest and re-conform. But that speed costs you twice in the end—once in artifacts, once in the redo you thought you'd avoided.

Frequently asked questions

Can I fix an IDT mismatch after grade?

Technically, yes — but it's a hostage negotiation with your own timeline. Once grading is locked, the color space transform is baked into every node, every power window, every second of shot matching. If you reassign the IDT now, the ACES transform recalculates from raw, but your grade was built on the wrong starting point. The odd part is—the fix usually works, but you'll be re-trimming every scene. I have seen a DI house spend three full days chasing the same skintones they already matched once. The catch? A grade applied on a mismatched IDT carries that original error through every subsequent transform. So yes, you can fix it. But the cost is time, and the risk is never quite getting back to the original look.

A better approach: catch the mismatch during the dailies review, not during the final color session. Most teams skip this—they assume the metadata is correct. Wrong order. You can check in the CDL or a quick RCM node before any creative work begins. That's where you save the pain.

Does a LUT hide the shift?

Not really — it just masks the symptom while the disease spreads. A LUT applied on top of a mismatched IDT doesn't correct the underlying color space math; it only bends the output to look acceptable in a single viewing condition. The problem is that LUTs are fixed, non-adaptive beasts. Switch from Rec.709 to P3? The LUT breaks. Deliver for HDR? The LUT clips. One concrete anecdote: a post house I worked with slapped a show LUT over an Alexa‑to‑ACEScg mismatch, and everything looked fine on the client monitor. Then the theatrical master needed a trim pass — and the seam blew out. Green fleshtones in shadow, cyan in highlights. That hurts.

What usually breaks first is the inter-format consistency. A LUT hides the shift inside one color space, but the moment you transcode to DCP, HDR10, or broadcast variants, the hidden error reveals itself. The fix is still the IDT — not a band‑aid LUT. Think of it like leveling a crooked table with a folded napkin. It works until someone bumps the table.

How do I check IDT metadata quickly?

Two ways, and neither requires a degree in color science. First: in your grading software, inspect the clip's source metadata — most tools expose the camera make, model, and color space tag. If your camera is an ARRI and the metadata reads 'REDWideGamutRGB', you have a problem. Second: load a waveform and a vectorscope on a neutral gray card or a MacBeth chart. If the gray isn't flat across R, G, and B (within 2–3%), and the vectorscope shows a consistent hue shift, the IDT is likely wrong. That's a five‑minute check, not a twenty‑minute science experiment.

The tricky bit is trusting what the metadata says. Sometimes the camera itself writes the wrong tag — firmware bugs, LUT applied in‑camera but not baked, or a DIT who accidentally set the wrong color space. Don't rely on the file header alone. Verify with a physical reference: a gray card shot on set. If the card in your timeline doesn't match the known values for that camera's color space, reassign the IDT. You can do this in Resolve's Color Management panel or in Baselight's Scene Settings. It's a two‑click fix — but only if you catch it before grade.

'We trusted the metadata. We shouldn't have. Twelve shots had to be re‑graded from scratch.'

— Senior Colorist, episodic television, after a 16‑hour conform session

So: manual reassign or rebuild?

When to choose manual reassign

Manual reassign works best when you're inside the DI timeline—locked, stressed, and staring at a shot that just won't neutralise. I've done this on two-hour turnarounds: you open the RCM, swap the IDT tag to a known-safe matrix like ACEScg, and pray the rest of the show holds. It's crude, but it moves. The catch is that you're faking a colour space—so if your source is Alexa LogC and you reassign to Rec.709, your primaries will rotate. Not a disaster for a single hero shot; a slow bleed across a 90-minute feature. Manual reassign also works when the original IDT metadata is simply missing or corrupted—common with proxies baked in Resolve 17-era workflows. One hard rule: never reassign without a side-by-side monitor showing a known reference. We fixed a 45-second spot this way last month—client never knew.

When to invest in custom IDT

Custom IDT is for when the mismatch lives in the camera itself—not the metadata. Think drone cameras, vintage glass conversions, or any sensor that emits a spectral signature no manufacturer bothered to target. Building a custom IDT means shooting a ColorChecker under controlled light, extracting the raw response, and baking a transform that maps that sensor into ACES primaries. It's not a weekend project—expect 2–5 days if your color scientist knows what they're doing. But here's the trade-off you don't read in white papers: once you build it, you own it. That transform becomes a deliverable. If the DP switches to a second body midway through production, your custom IDT may not hold. Most teams skip this until they see the skin tones clip—by then, the show's already conformed. Verdict: invest custom if the show has fewer than 15 cameras and a budget that can absorb a mid-week scientist rental.

I've seen a manual reassign buy a day for a grade that then broke in the final render—and a custom IDT save a six-camera show at the cost of two lost prep days.

— colourist who learned both lessons the hard way, 2024

The one step you should never skip

Before you choose either path, stop. Validate the baseline. Pull three frames: one from a neutral grey chart, one from a skin tone region, one from a saturated primary. Pipe them into a viewer that shows raw R,G,B excursions—not just a waveform. If your reassign shifts the grey point by more than 2% in any channel, manual reassign is lying to you. If your custom IDT produces a flat curve on the same grey, it's been overfitted. I have seen teams rush into a rebuild only to discover the original IDT was fine—the DP had simply flagged the LUT. So check. It takes seven minutes. That seven minutes has saved me from re-grading 18 shots more than once. The wrong choice here isn't theoretical—it's a day of re-conform, a frustrated client, and a colour shift that follows you to the final DCP. Don't guess. Measure.