Automatic Shot Matching for Indie Filmmaking | Leumos AI

Automatic shot matching for indie filmmaking solves the single biggest time-sink in a low-budget grade: equalizing 400 to 2,000 clips shot across multiple camera bodies, lighting setups, and shoot days. The right AI tool turns the first-pass equalization into a review pass, leaving the colorist-editor free for actual creative decisions inside DaVinci Resolve.

I've been a colourist for four years — DaVinci Resolve Certified, BFA in cinematography — and the projects that nearly broke me weren't the $200K branded jobs. They were the indies. A 14-minute short shot on a BMPCC 6K and a borrowed Canon C70, intercut every other shot. A 78-minute first feature mixing FX6 A-cam with a RED Komodo B-cam because the budget made you do what the budget made you do. Every one of those projects involved me sitting in a dark room for an entire weekend running node soup on shots that should have been one Copy-Paste-Grade apart. I built Leumos because I was tired of losing weekends to math the computer should be doing.

The Math of Indie Multi-Cam Is Brutal

Indie sets don't get to pick one camera body. You get the BMPCC 6K the DP brings, a C70 the production company already owns, an FX6 the first AC has on rent, and maybe a Komodo for the day someone called in a favour. By the end of principal photography you have BRAW from the Pocket, ProRes RAW from the Sony, C-Log3 from the Canon, and possibly REDCODE RAW from the Komodo — four different log gammas, four sensor color science chips, four readouts that respond to skin tones differently.

That's before lighting drift. A 22-day shoot has weather changes, gel swaps, practicals burning out, magic-hour scenes that were supposed to be middays. Even if you had one camera body, you'd have multiple looks coming out of the same lens.

The math gets worse. An 80-minute feature at an average shot length of four seconds is roughly 1,200 cuts. Each cut needs to feel like it belongs to the same scene, same world, same emotional register. Manual shot matching in DaVinci Resolve — node per clip, eye-dropper, scopes, gallery stills — is realistically a 40 to 60 hour pass before you've made a single creative choice.

On a $20K budget, that's not a luxury. That's the whole grade. You'll burn the entire post window equalizing footage and ship a competent neutral pass to the festival when you wanted something that punched above its budget. The bottleneck isn't your taste. The bottleneck is the time.

What Manual Shot Matching Costs You Creatively

Think about Linus Sandgren's work on La La Land. The Technicolor palette — magenta dress against teal sky, the gold-lavender freeway sequence — is internally consistent because Linus and his colorist had budget, time, and a 35mm photochemical reference to chase. Indie features don't have that. You have a hard drive, a producer texting about a deadline, and the lingering memory of Hong Kyung-pyo's tonal control in Parasite — that cold basement gray cutting against the warm Park house interior — and you wonder how you're supposed to land within shouting distance of that on a coffee-and-rent budget.

The cost isn't just hours. It's creative tax. By the time you've equalized 800 clips manually, you have nothing left in the tank for the actual look. You ship a technically correct grade instead of a memorable one. The festival cut goes out at 80% of what you imagined when you were storyboarding the look book in pre-production.

I've watched friends finish their feature, look at the timeline, and realize they spent every available hour on equalization and zero hours on the thing that would have made the grade actually theirs. Festival programmers don't reward correct. They reward voice. And voice requires hours you don't have if you're spending them all on color-balance math.

If you're an indie filmmaker, we're building this for you. Leumos AI is in closed beta — join the early-access list and you'll be in the first 500 (50% off for life).

How AI Shot Matching Actually Works Under the Hood

I'm building Leumos around three layers that mirror how a colorist already thinks.

First, you transform out of log. Input Color Space LUT takes whatever you shot — S-Log3, C-Log3, BMDFilm, V-Log — and converts it to Rec.709 in one click. This is not a creative LUT. It's a math operation that puts every camera into the same color space so the algorithm can compare them apples to apples.

Second, AI Scene Cut Detection chops your upload into discrete shots on a timeline with thumbnails. You don't manually trim clips, you don't drag in-and-out points. You drop a 14-minute short and you get a shot-by-shot timeline ready to grade. For the occasional cross-dissolve or whip-pan the AI misreads, the Manual Cut Tool splits it with a single click.

Third, Match All runs an equalization pass across the whole timeline — exposure, contrast, brightness, saturation, hue. The algorithm doesn't pick a single hero shot. It looks at the median of your footage and pulls outliers toward the center. That's the equivalent of the opening balance pass of any indie grade collapsed into one button.

If you want a specific look — Bradford Young's blue-amber palette from Arrival, or the cool desaturated melancholy of Chayse Irvin's grade on Aftersun — drop a still into Reference Image Grading and pull the intensity slider from 0 to 100%. The footage moves toward the reference. The point isn't that this replaces a colorist's eye. The point is that the colorist's eye should be deciding which direction to push, not eye-droppering skin tones for the 400th time in a single weekend.

Where AI Shot Matching Breaks (Be Honest)

I won't sell you magic. AI shot matching does three things badly right now, and I'd rather you know going in than feel oversold after.

Mixed lighting inside a single frame — a window blowing out behind a tungsten-lit close-up, or a streetlamp competing with a neon practical — still confuses the algorithm. Match All will equalize the overall exposure but the white balance will fight itself. You'll still need Manual Primaries to push temperature and tint per-shot, especially in reception scenes, club scenes, or any practical-heavy interior.

Skin tones across different ethnicities under aggressive practicals are a judgment call no AI handles cleanly today. Match All gets you into the neighbourhood. The rest is you, by hand, every time.

Day-for-night, brand-Pantone compliance, and high-contrast genre looks — Fincher-level desaturation, the brutal contrast of recent A24 horror — are creative decisions. AI can match references. It cannot invent. You're still the colorist on those calls.

If anyone tells you their AI tool handles all of this on autopilot, ask to see the raw footage and the final grade side by side. I've tested every tool on the market over the last twelve months and none of them have fully solved these three problems. Including the one I'm building.

The Resolve-Friendly Workflow I'm Building For

The colorist-editors I know don't want to leave Resolve. They want a faster first pass and they want to finish where the toolset is mature.

When Leumos launches, the workflow will look like this:

1. Cut your indie in Resolve and export a 1080p ProRes proxy of the assembly
2. Drop the proxy into Leumos in the browser — no install, no plugins
3. AI Scene Cut Detection builds the shot-by-shot timeline
4. Input Color Space LUT handles your camera's log gamma
5. Match All runs the equalization pass across the full timeline
6. Reference Image Grading lands the creative direction from a film still or reference frame
7. Layer a creative LUT from the Preset LUT Library with the intensity slider if you want a baseline look
8. Export the graded render and conform back into Resolve for final finish, scopes, secondaries, and delivery

You're not replacing Resolve. You're skipping the clip-by-clip equalization pass and arriving at the creative grade with the canvas already prepped. That's the workflow I wanted when I was the colorist-editor on my own short and lost three consecutive weekends to node soup. It's the workflow I think indie features deserve when the budget is $15K and the festival deadline is six weeks out.

I'd rather you spend those weekends shaping a look you're proud of than equalizing exposure between a BMPCC 6K wide and a C70 close-up. The product is in closed beta. The first 500 signups on the early-access list lock in 50% off for life. If you're cutting an indie this year and the thought of another node-soup weekend makes you want to quit color entirely, this is built for you.

Frequently asked questions

How long does automatic shot matching take on an 80-minute indie feature?

A typical 80-minute feature with around 1,200 cuts runs through AI Scene Cut Detection and Match All as automated passes rather than a clip-by-clip session. Compare that to a manual first-pass equalization in DaVinci Resolve, which realistically takes 40 to 60 hours across a multi-cam BMPCC/C70/FX6 shoot. The AI pass doesn't replace your creative grade. It just gets you to the starting line of the creative grade without the clip-by-clip pass.

Does AI shot matching work with footage from BMPCC, Sony, and Canon bodies in the same timeline?

Yes, but you'll want to flatten to a common color space first — and to export out of the raw containers before upload. Input Color Space LUT converts S-Log3, C-Log3, BMDFilm, V-Log and similar log gammas to Rec.709 in one click, putting every camera body in the same playground. After the log transform, Match All treats the footage as one timeline regardless of origin. The algorithm doesn't care whether a frame came from a Komodo or an FX6 once it's in 709, which is exactly the point. You stop fighting the differences between sensor chips and start grading the picture.

Can I round-trip my Leumos grade back into DaVinci Resolve for final finish?

Leumos returns a rendered, graded file rather than grade data, so the honest version of a round trip is: use the render directly where it meets delivery, or treat it as the approved reference your Resolve finishing pass matches on the conform. Exporting grades into Resolve nodes is not something the closed beta does. You're using Leumos for the equalization and reference pass, not for QC-level final delivery. Resolve is still the finishing room.

Will AI shot matching get skin tones right across different lighting setups?

Honestly, 70-80% of the way there. Skin tones under controlled lighting equalize well — daylight exteriors, soft-box interiors, anything with a consistent key. Where it struggles is mixed practicals: a tungsten table lamp competing with a daylight window, neon-heavy nightlife scenes, or a wide range of skin tones inside a single sequence. You'll still need Manual Primaries for those shots. If a vendor promises perfect skin under all lighting on autopilot, they're either lying or they haven't tested it on real indie footage.

What's the difference between Match All and Reference Image Grading?

Match All is internal consistency — it equalizes every shot in your timeline against the median of your own footage. Nothing creative, just neutral cohesion. Reference Image Grading is external direction — you drop a still from a film, a photographer, or a look book and the AI pulls your footage toward that look with an intensity slider. The typical workflow is Match All first for cohesion, Reference Image second for creative target. Two different jobs. Both useful. Neither one replaces the other and they're meant to stack.

Is this a DaVinci Resolve replacement?

No, and I'd be lying if I said it was. Resolve is the most mature color tool on the planet. Leumos is built to replace the first-pass equalization step that nobody enjoys — the part where you're not making creative decisions, just normalizing math across camera bodies and shoot days. You finish in Resolve. The pitch is Leumos for the first pass, Resolve for the finish, not Leumos instead of Resolve. If you love node trees and scopes, you'll still spend most of your finish time in Resolve. You'll just get there with the timeline already equalized.

How does the pricing work for an indie feature with a 4-week post window?

Plan structure and pricing have not been announced yet. The first 500 waitlist signups lock in 50% off for life.

---

Leumos AI is in closed beta. The first 500 waitlist signups lock in 50% off for life. Join the early-access list →