Look & Composition

What Is Film Grain?

Also called grain, grain structure, film emulation

Film grain is the visible texture produced by clumps of light-sensitive silver crystals in photographic emulsion. Faster stocks have larger crystals and coarser grain. Digital sensors have no grain at all, so the look is now added deliberately to soften digital cleanliness and to signal a period or a format.

Where the texture comes from

Photographic emulsion is a layer of silver halide crystals suspended in gelatin. Light converts individual crystals, and development clumps them into visible particles. The size of those crystals is a design tradeoff: bigger crystals catch light faster, so a high-speed stock is sensitive and coarse, while a slow stock is fine and needs more light. That single tradeoff is why grain has always carried meaning. Coarse texture said available light, night, news, urgency. Fine texture said control, budget, daylight.

Digital sensors have no equivalent structure. What they produce instead is noise, and the difference is not cosmetic. Noise appears in the shadows, carries colour speckle, and often repeats in a fixed pattern that the eye recognises as an error. Grain sits across the whole image, is mostly luminance, and is different in every single frame. That per-frame randomness is the part people forget, and it is the part that matters most.

Why clean footage gets grain added back

  • Masking banding. Smooth gradients in skies and walls break into visible steps at delivery bitrates. A light layer of grain dithers the boundaries away.
  • Covering artificial smoothness. Upscaled, denoised, and generated images all share a plasticky evenness. Grain gives the surface something to be.
  • Unifying mismatched sources. Shots from different cameras, stock footage, and generated inserts sit together far more comfortably under a single grain layer.
  • Era and format signalling. Coarse grain plus reduced saturation reads as archival immediately, with no other cue needed.
  • Perceived sharpness. Texture in the mid-tones makes an image feel more detailed than the same image scrubbed clean, even though it measures worse.

Getting it in generated images and video

For stills, the prompt route works reasonably well, because grain is texture and models render texture confidently. Name the stock or the format rather than asking for grain in the abstract.

  • ✅ shot on 35mm Kodak stock, visible fine grain, slightly muted colour
  • ✅ 16mm handheld documentary look, coarse grain, cool cast
  • ⚠️ grainy often produces heavy uniform speckle that looks like sensor noise
  • ⚠️ vintage gets you grain plus scratches, vignetting and colour shifts you may not want

For video, be sceptical of asking the model at all. Generated grain is frequently frozen in place across the clip, so the texture sticks to the frame while the picture moves underneath it, which reads as dirty glass rather than film. The reliable approach is to generate clean and add grain in an editor, where the plate is regenerated every frame and you control the size and strength.

One consequence worth planning for: grain is expensive in compression terms. It is random detail, which is exactly what codecs cannot predict, so a heavily grained master needs a noticeably higher bitrate to survive upload. If a clip looks fine locally and turns to mush after publishing, the grain layer is the first suspect.

Checks worth doing

Look at a flat area at full size, not a face. Grain should be even across the frame rather than concentrated in the shadows, and it should not have a visible repeating tile. In video, step through three consecutive frames: if the speckle pattern is identical, it is static and should be replaced. Finally, view at the size the audience will use. Grain judged on a 200 percent zoom is almost always applied far too strong.

The prompt for this

A starting point that reliably produces the effect. Adjust the subject and setting; keep the technical clauses.

Shot on 35mm Kodak stock, visible fine film grain, slightly muted colour, a diner interior at night lit by a single overhead fluorescent, 40mm

Try Film Grain yourself

Open the generator with a starting point already filled in.

AI Video GeneratorAI Image Generator

Frequently asked questions

What is the difference between film grain and digital noise?
Grain is physical structure in the emulsion, so it is irregular in shape, roughly luminance-only, and different on every frame. Digital noise comes from the sensor and readout, so it carries coloured speckle, often shows a fixed pattern, and clusters in the shadows. Grain reads as texture, noise reads as a fault.
Does adding grain reduce image quality?
It reduces measured sharpness and raises the bitrate you need, but it can raise perceived quality. Grain masks banding in gradients, hides the plastic smoothness of upscaled or heavily denoised footage, and gives the eye texture to hold on to.
How much grain is too much?
When you notice it before you notice the image. On a delivery viewed at normal distance, grain should be visible in flat mid-tones and invisible on a face. If it crawls in the sky or looks like dust on a lens, it is too coarse rather than too strong.
Where in the pipeline should grain go?
Last. Add it after upscaling, denoising and colour grading, because every one of those steps will smear or amplify grain that is already baked in. Grain applied before an upscale stops looking like grain and starts looking like texture on a wall.
Why does grain look wrong in AI-generated video?
Because models paint it as a static texture rather than regenerating it per frame. Real grain is different in every frame, which the eye reads as air. Frozen grain reads as dirt on the lens, and it is the fastest way to make a generated clip feel synthetic.

Related terms

Back to Look & Composition