16mm Found Footage — Analog Grain & Black-and-White Look

Also goes by Deep Woods Footage. A style LoRA for MiniMax H3 that renders footage the way analog 16 mm black-and-white film looks: real film grain, crushed but readable shadows, halation around lights, soft corners — the grainy, high-contrast register of late-90s expedition horror.

Trigger: foundfootage16 — start every prompt with it.

What it does

- Night: a single flashlight beam in pitch-black woods, light falling off fast into pure black, a tent

glowing from inside, figures half swallowed by the dark

- Daylight: muted, grain-heavy monochrome — overcast woods, grey skies, mud, leaf litter, creeks

- Vibe: handheld, unstaged, documentary. Cramped framing, POV shots, looking down at the forest floor while walking

- Texture: the grain is not simulated noise — it was extracted from real film scans and transplanted, so it clumps the way silver-halide grain does. The optics are soft (16 mm, not digital): no micro-contrast crispness, but deep focus — no bokeh, no shallow depth of field, no background blur.

Recommended strength

Scene - Strength

Daylight / bright interiors - 0.8

Night, flashlight, deep blacks - 0.9 – 1.0

Strength scales the whole look (grain, contrast, halation). Above ~1.1 the grain starts to break up between frames and can read like drizzle. Keep brightness in the prompt instead ("a flashlight beam lighting the path and the nearest trees" — not "brighter exposure", which flattens the falloff).

Keeping the grain crisp: grain is the first thing a video encoder throws away. If your export looks

smoother than the samples, encode at CRF ≤ 16 with -tune grain, or export the frames and encode them

yourself. On-site generation applies its own encoding — expect softer grain there.

Examples

```

foundfootage16, a person walking away down a muddy forest trail at night, a flashlight beam lighting the path and the nearest trees, high contrast, deep blacks, exposure falls off quickly away from the beam, soft 16mm optics, deep focus, no background blur

foundfootage16, two people in winter jackets walking along a muddy forest path in overcast daylight, handheld camera behind them, slight camera shake, soft 16mm optics, grainy black-and-white foundfootage16, a small dome tent glowing from inside in dark woods at night, branches in the foreground, high contrast, deep blacks

```

Natural sentences work better than keyword lists. Useful words: soft 16mm optics, deep focus,

exposure falls off quickly, handheld. Avoid bokeh, shallow depth of field, out of focus (they

fight the deep-focus look) and everything sharp (it pushes the model toward digital crispness).

Training: 165 images — 120 synthetic renders of forest/night motifs (degraded with real film grain

transplanted from scans, soft 16 mm optics) plus 45 hand-curated frames from analog 16 mm black-and-white footage; natural captions. MiniMax H3, one-frame mode, guidance-loss recipe, rank 16, 3500 steps.

Category: Style.

Notes

- SFW.

- Trained video-only: generated audio is untouched by this LoRA.

- Camera movement is not baked in — pair it with a handheld/shake camera-control LoRA for the wobble.

- The grain is strong by design. Because the LoRA was trained on stills, heavy grain can shimmer slightly from frame to frame; if that bothers you, drop the strength a notch rather than removing the grain.

- Please don't use it to recreate identifiable real people, actual events, or existing footage

shot-for-shot.