LUTs, Film Grain, and the Film Look for AI Video

The best LUTs for AI video are film-stock emulations applied at reduced strength — not stylized "cinematic" looks at 100%. Start with the free Kodak 2383 and Fuji 3510 print emulations at 60–80% opacity, or paid options like FilmConvert Nitrate ($99–$199) and Dehancer ($149/year). Apply them in a fixed order: upscale, color correct, LUT at 50–70%, 35mm grain at 8–15%, then a light blur.

Why standard film LUTs fight AI footage

AI-generated footage arrives with a signature problem: it is too clean. Output from current video models comes back ultra-sharp, with plasticky skin texture and contrast already baked in — one documented production noted that its generated clips had "an ultra-sharp, plasticky skin quality that must be corrected in post to achieve realism."

That breaks the assumption most film LUTs are built on. Conversion and look LUTs are designed for flat, low-contrast log footage from a camera sensor. Feed them an AI clip that is already saturated and contrast-heavy and the LUT double-applies: shadows crush, highlights clip, and skin tones shift into orange or magenta. Stylized LUT packs make it worse — you are stacking a strong look on top of footage that is already a rendered interpretation of a look.

The fix is twofold: choose emulation LUTs (which model the subtle behavior of film print stock — highlight rolloff, gentle color crosstalk — rather than imposing a heavy grade), and apply them at reduced opacity so the LUT shapes the image instead of overwriting it.

The best LUTs for AI video, by tier

The options that consistently work on AI footage fall into three tiers: free film-print emulations, paid stock-emulation plugins, and a LUT you generate yourself from one hero clip.

Free film-stock emulations to start with

Kodak 2383 and Fuji 3510 are film print emulation LUTs that ship free inside DaVinci Resolve (under Film Looks in the LUT browser). They emulate the print stage of the photochemical pipeline — the final stock a movie was actually projected on — which is exactly the layer of subtlety AI footage lacks. Apply either at 60–80% opacity over a corrected clip. Kodak 2383 leans warmer with teal-shifted shadows; Fuji 3510 runs slightly cooler and softer. Test both on one hero shot before committing.

Paid LUTs and plugins worth the spend

FilmConvert Nitrate ($99–$199) models specific named film stocks and pairs each with grain scaled correctly to your delivery resolution, so the stock and the grain stay matched instead of fighting. Dehancer ($149/year) goes further than color: it adds halation, bloom, and gate weave — the optical and mechanical artifacts of real film that AI footage has none of. Those artifacts are what reads as "shot on film" at a subconscious level, which makes Dehancer particularly effective at disguising the digital flatness of generated clips. Both plugins include strength controls — keep them in the same 50–70% range as a standalone LUT.

Generate your own LUT from a hero clip

Grade one representative shot manually until it looks exactly right — balance, contrast curve, palette — then export that grade as a .cube LUT (in Resolve: right-click the graded clip's thumbnail → Generate LUT → 33 Point Cube) and apply it across the whole project at reduced strength. This anchors every clip to one approved look, which matters more on AI projects than on camera footage: generated clips drift in color from generation to generation, and a hero-clip LUT pulls them back toward a single reference. It is the grading equivalent of locking reference assets before production begins.

The film-look pipeline: correct, LUT, grain, blur

The LUT is one layer in a sequence, and the order is what makes it work. Documented productions run the film-look pass like this:

1. Upscale first. One production ran Topaz Astra on invideo as the first step of its post pipeline, before any color work. Grain and blur added at delivery resolution stay correct; grain added before an upscale gets stretched into mush.
2. Color correct per clip. Balance exposure and white balance on each clip individually so the LUT receives consistent input. AI clips from the same scene can vary in temperature and contrast — correction is what lets one LUT serve the whole timeline.
3. Apply the LUT at 50–70% strength. Full strength on AI footage produces the crushed, over-cooked result described above. Reduced opacity keeps the emulation's character without destroying the image underneath.
4. Add 35mm film grain at 8–15% opacity. Grain breaks up the over-clean digital surface and adds the texture the model never rendered. Match grain size to your delivery resolution.
5. Finish with a light blur. A small amount of softening — well under one pixel of Gaussian blur, or a subtle mist effect — takes the artificial edge sharpness off. The principle from documented production work: adding a small amount of blur, grain, and color grading on top of AI-generated footage moves it closer to live-action film.

Each step is small on its own. Stacked in order, they convert the telltale AI sheen into something that reads as photographed.

How to apply it in DaVinci Resolve and Premiere

DaVinci Resolve: build a serial node chain on each clip — Node 1 for correction (exposure, white balance), Node 2 for the LUT with the node's key output gain set to 0.5–0.7 to control LUT strength, Node 3 for the Film Grain OFX plugin (35mm preset, opacity 8–15%), Node 4 for a slight Gaussian blur or mist from the Blur palette. To apply the look project-wide, put the LUT, grain, and blur nodes at the timeline or group post-clip level and keep only per-clip correction on individual clips.

Premiere Pro: correct each clip in the Lumetri Basic tab, load the LUT in the Creative tab and pull the Intensity slider to 50–70, overlay a 35mm grain plate (or a grain plugin) on a track above set to Overlay or Soft Light at 8–15% opacity, then add a Gaussian blur of roughly 0.5–1.0 on an adjustment layer over the whole timeline.

Either way, the structure is identical: correction is per-clip, the film look is global.

Win the film look before the LUT — at generation

The less the LUT has to rescue, the better it performs — so specify the look in the generation prompt itself. Documented productions bake three things into every prompt: the lighting source, the palette, and a film or cinematographer attribution. One production encoded a director's lighting grammar as an explicit 85:15 dark-to-light ratio in its prompt language. Lighting prompts work best when they name the source: "warm yellow from the lamps only" produces more accurate results than generic "warm lighting."

One production formalized this as a 9-element prompt assembly order — camera spec, lens, lighting source, palette, composition, atmosphere, mood, film/DP attribution, negative prompt — held across every frame of the project, and had its agent define an 8-step color grading guidance process alongside it. Inside invideo — an agentic video creation tool with all the current video models (Veo, Kling, Seedance 2.0) and upscalers available — the invideo agent holds these style directives in context across every shot, so clips arrive pre-matched in palette and contrast and the LUT layer stays light. If you systematize the full look in a treatment document the invideo agent keeps loaded, every generation inherits it without re-prompting.

Footage that is generated in-palette needs a 60% LUT and 10% grain. Footage generated without color direction needs a rescue grade — and rescue grades are where AI video starts looking processed.

What the film-look layer costs in context

The entire film-look layer is one of the cheapest line items in an AI production. The free Kodak 2383 and Fuji 3510 emulations cost nothing; Dehancer runs $149/year and FilmConvert Nitrate $99–$199 one-time; upscaling runs on invideo credits as part of the normal pipeline.

Set that against documented all-in production budgets — including a ~$950 animated episode — which work out to $315–$750 per finished minute. A $150 plugin amortized across multiple projects adds low single-digit percentage points to any of those budgets, and the free tier adds zero. The film look is not where AI video costs accumulate — generation iteration is. Spend the money there and grade with what ships free in Resolve until a specific project justifies the plugins.

FAQ

What opacity should I apply a LUT to AI video?

50–70% for most film LUTs, and 60–80% for the free Kodak 2383 and Fuji 3510 print emulations, which are gentler by design. AI footage already carries baked-in contrast and saturation, so full-strength application crushes shadows and clips highlights. Set the strength via node key output gain in Resolve or the Lumetri Intensity slider in Premiere.

Does film grain make AI video look more realistic?

Yes — 35mm grain at 8–15% opacity breaks up the over-clean, over-sharp surface that marks footage as AI-generated. Documented production work treats a small amount of grain, blur, and color grading as the standard pass that moves AI footage closer to live action. Keep it subtle: visible grain reads as an effect, low-opacity grain reads as film.

Should I upscale before or after grading?

Upscale first. One documented production ran Topaz Astra on invideo as the first step of its post pipeline, before any color work, so the LUT, grain, and blur are applied at final delivery resolution. Grain added before an upscale gets stretched and loses its texture.

Can I use one LUT across all my AI-generated clips?

Yes, and you should — a single LUT is what unifies clips that drift in color between generations. The requirement is per-clip correction first: balance exposure and white balance on each clip so the LUT receives consistent input. Generating your own LUT from one manually graded hero clip is the strongest version of this approach.

Sources

• DaVinci Resolve — Blackmagic Design — Resolve ships the Kodak 2383 and Fuji 3510 film print emulation LUTs free, plus the Film Grain OFX plugin referenced in the Resolve workflow.
• FilmConvert — pricing and stock-emulation details for FilmConvert Nitrate.
• Dehancer — pricing and feature set (halation, bloom, gate weave) for the Dehancer film emulation plugin.

Watch these to see the techniques in action: