Cinematic Video Prompt Keywords: Aspect, Grain, Lens

Cinematic video prompts fail when you paste still-image adjectives into a motion box. Kling AI, Runway Gen-4, Luma Dream Machine, and Pika treat "cinematic 8K masterpiece" as a pretty poster with random camera drift. Fix the look with three keyword layers: aspect ratio language matched to your output lane, film grain that adds texture without smearing motion, and lens cues that tell the renderer how wide or tight the frame feels. This guide is a keyword reference for video hosts as of August 2026. You leave with paste-ready aspect, grain, and lens blocks, model dialect notes, stack order on top of subject and motion lines, and a soft path to PromptMake /image when your clip starts from a hero still.

What cinematic video prompts need beyond still keywords

Still-image guides teach lighting ratios, Kelvin temps, and chiaroscuro for one frozen frame. Video hosts need those words plus temporal discipline. A lens keyword that works on a Midjourney still can fight motion blur when the subject walks across frame. Grain that reads as mood on a portrait can turn a product spin into muddy pixels on a five-second Kling clip.

Cinematic video prompts name look cues that survive seconds of change. Aspect ratio tells the model which canvas to protect. Grain tells it how much analog texture to add per frame. Lens language tells it field of view and depth behavior while the camera moves. You write subject and motion first using any SMCD workflow you already use. This page fills the look layer that sits after those pillars.

Who should lean in: creators cutting short hero clips, marketers testing widescreen teasers, designers storyboarding motion before a longer edit. Who can skim: teams that only export stills and never open a video UI. This article does not repeat lighting keyword catalogs from still-image posts. It stays on aspect, grain, and lens for motion hosts.

Aspect ratio keywords for video hosts

Aspect ratio is the first cinematic keyword because it sets the frame before grain or lens cues apply. Video UIs expose ratio pickers on Kling, Runway, Luma, and Pika. Prompt text should repeat the same ratio so the model does not invent letterboxing or crop your subject mid-clip. Write the ratio as a fraction, a named format, or a platform lane.

Weak ratio language: "cinematic widescreen." Strong ratio language: "16:9 horizontal," "2.39:1 anamorphic widescreen," "9:16 vertical portrait frame." Pair ratio with one composition hint: "subject centered in 16:9," "figure in lower third of 9:16 frame." That pairing stops hosts from placing a portrait subject in dead center of a vertical lane when you wanted headroom for captions.

Match prompt ratio to UI ratio every time. A 16:9 prompt pasted into a 9:16 slider forces the model to invent top and bottom space. Crop your hero still to the target ratio before image-to-video upload. Ratio keywords in text plus cropped upload beat ratio keywords alone.

16:9, 21:9, and 2.39:1 widescreen language

16:9 is the default horizontal lane for YouTube, web hero, and most Runway Gen-4 exports. Keywords: 16:9 aspect ratio, horizontal widescreen, 1.78:1, HD cinematic frame. Paste example for Kling text-to-video: "16:9 horizontal frame, medium shot of a yellow taxi on wet asphalt at dusk, slow dolly in only, 5 seconds."

21:9 ultrawide reads as premium cinema and monitor-fill B-roll. Keywords: 21:9 ultrawide, 2.35:1 cinematic aspect, scope frame, letterbox cinematic format. Use when you want empty side space for titles in post. Avoid 21:9 on hosts that only export 16:9 without crop; check the export menu first.

2.39:1 anamorphic scope is the keyword stack for true film-width feel. Keywords: 2.39:1 anamorphic, scope aspect ratio, cinemascope frame, wide cinematic letterbox look. Runway Gen-4 and Kling both respond to explicit 2.39:1 language in mid-2026 public docs. Pair with lens cues from the lens section: anamorphic lens flare, oval bokeh, horizontal lens streaks on highlights. Do not ask for scope ratio and 9:16 vertical in the same prompt.

9:16 and 1:1 vertical and square lanes

9:16 vertical owns TikTok, Reels, Shorts, and Stories. Keywords: 9:16 vertical video, portrait aspect ratio, mobile full-screen frame, vertical cinematic clip. Write subject placement for vertical: "figure in center third, headroom for text overlay," or "product on lower half, negative space above." Vertical lanes punish horizontal composition language like "wide establishing shot" unless you mean a tall narrow street canyon.

1:1 square fits feed posts and product loops. Keywords: 1:1 square frame, square cinematic clip, centered subject in square format. Square clips need tighter subject anchors because horizontal breathing room disappears. One primary subject, one motion beat, one camera path.

Pika short-form lanes often default to vertical or square as of August 2026. Confirm the ratio picker before paste. Luma Dream Machine accepts 16:9 and 9:16 on most consumer tiers. Name the ratio in the first clause so later grain and lens keywords attach to the correct canvas.

Film grain and texture keywords

Film grain adds analog texture frame to frame. In video prompts, grain is a density dial, not a mood word. Light grain sells cinema without hiding subject edges. Heavy grain can swallow fine detail when the camera pushes in or when fabric moves.

Video models interpret grain differently from still hosts. Motion already introduces inter-frame noise. Stacking "heavy grain, ultra noisy, gritty texture, VHS damage, film scratches, dust specks, light leaks" in one Kling pass often produces flicker or muddy faces. Pick one grain tier per clip. Add a second grain adjective only after a clean base render.

Grain keywords belong after subject, motion, camera, and ratio lines. They modify look. They do not replace motion verbs. "Beautiful cinematic grain" without a subject noun still returns random drift.

Fine grain vs heavy grain

Fine grain keywords: subtle 35mm film grain, light photochemical grain, fine Kodak grain, gentle film texture, soft analog grain overlay. Use on portraits, product hero clips, and image-to-video uploads where face identity must hold. Fine grain pairs with shallow depth and warm practicals.

Medium grain keywords: visible 16mm film grain, documentary film texture, moderate grain throughout, shot on 16mm aesthetic. Use on street scenes, handheld B-roll, and lifestyle clips where slight texture helps and perfect skin pores matter less.

Heavy grain keywords: heavy 8mm film grain, coarse grain, high ISO noise aesthetic, gritty analog texture, VHS-style noise (use alone, not stacked with heavy 8mm). Reserve heavy grain for night exteriors, music-video nods, and deliberate retro lanes. Test heavy grain on three-second clips before you commit credits to a ten-second Runway pass.

Negative guardrails when grain overruns: some hosts accept "clean digital image, no film grain" on retry. Runway and Luma respond to that contrast when the first pass flickered. Do not rely on negatives alone; reduce grain adjectives first.

When grain fights motion clarity

Grain plus fast subject motion equals smear on consumer video tiers. A runner sprinting across frame with heavy 8mm grain often loses leg detail. Fix: drop to fine grain or lock camera and reduce subject speed in the motion line.

Grain plus push-in camera motion magnifies texture toward the end of the clip. A slow dolly in on a face with medium grain can look fine at start and noisy at finish. Fix: fine grain on push-ins, or locked camera with medium grain on static subject motion.

Image-to-video uploads already carry sensor noise from the source still. Add fine grain only unless you want a deliberate film override. "Preserve reference, add subtle 35mm grain" beats "heavy grain" on Luma portrait uploads.

Product spins and label reads need minimal grain. Type on a bottle label disappears under coarse grain. Use clean digital product shot, no visible grain or omit grain entirely on packshot clips.

Lens and focal length keywords

Lens keywords tell the video host how wide or tight the virtual camera sees the scene. They differ from camera movement words. "Slow dolly in" is movement. "35mm spherical lens" is field of view. Write both, but keep them in separate clauses so Kling and Pika do not merge them into a single mushy "cinematic move."

Focal length numbers anchor perspective. Ultra-wide exaggerates depth and edge stretch. Normal lenses feel human-eye neutral. Telephoto compresses background toward subject. Anamorphic adds oval bokeh and horizontal flare on highlights. Name one lens family per clip unless you are emulating a deliberate lens change shot, which most short hosts handle poorly in one pass.

Pair lens keywords with shot size: 24mm wide establishing, 50mm medium shot, 85mm close-up portrait, 100mm telephoto compression. Shot size plus millimeter beats millimeter alone.

Wide angle and anamorphic cues

Wide angle keywords: 24mm wide angle, 28mm lens, wide establishing lens, expansive field of view, slight wide-angle distortion at frame edges. Use for interiors, landscapes, and street establishing clips. Wide lenses on 9:16 vertical can feel immersive when you name low angle 24mm vertical frame.

Anamorphic keywords: anamorphic lens, 2x anamorphic squeeze, oval bokeh, horizontal blue lens flare on highlights, anamorphic streak flare, scope lens character. Runway Gen-4 and Kling read anamorphic language when ratio is also scope-wide. Stack: "2.39:1 scope frame, anamorphic lens, oval bokeh, subtle horizontal flare on neon sign." One flare source per clip.

Ultra-wide caution: 14mm ultra-wide, fisheye distortion on close portraits produces facial stretch in video. Reserve ultra-wide for exteriors, architecture, and action at distance. If faces warp, step up to 35mm and regenerate.

Pika short clips respond to simple lens tags: wide cinematic lens, anamorphic look. Keep Pika lens lines short. Long lens stacks compete with duration limits on sub-five-second lanes.

Telephoto compression and shallow depth

Telephoto keywords: 85mm portrait lens, 100mm telephoto, 135mm compressed background, telephoto lens compression, subject isolated with long lens. Compression pulls background toward subject. Use for portrait inserts, product hero isolation, and "busy street blurred into bokeh balls behind subject."

Shallow depth keywords: shallow depth of field, f/1.4 aperture feel, creamy bokeh background, subject sharp background soft, rack focus (only when you want focus shift as the motion beat). Rack focus is a motion event. Write timing: "0–2s face sharp, 2–4s focus shifts to hand."

Macro and product: macro lens close-up, 100mm macro, extreme shallow depth on label. Pair with locked camera on short clips so focus plane stays stable. Macro plus dolly plus grain is a common failure stack on Hailuo three-second tests.

Avoid conflicting lens asks: "24mm wide" and "85mm portrait compression" in one sentence split the difference into a muddy mid lens. Pick one focal length family per generation pass.

How to stack aspect, grain, and lens with motion

Stack order matters. Subject and motion come first. Camera movement second. Aspect ratio third. Lens fourth. Grain last. Duration lives in the UI and in your beat list, not in a grain clause. This order mirrors how Kling, Runway, Luma, and Pika parse prompts in public examples as of mid-2026.

Example stack for a six-second Runway Gen-4 street clip: Subject and motion in line one. Camera in line two. Aspect and lens in line three. Grain in line four. Full paste: "Low wide shot of wet asphalt crosswalk at dusk, yellow taxi approaches slowly for three seconds then brake lights brighten, static tripod camera, 16:9 frame, 24mm anamorphic lens with oval bokeh and subtle horizontal flare on neon, fine 35mm film grain."

Cut look keywords before you cut motion when a render fails. If the clip drifted, fix subject and motion first. If the clip held motion but looks like a phone video, add lens and ratio on retry two. Add grain on retry three.

Step-by-step paste order

1. Write subject noun, material, and place.
2. Write timed motion beats matched to UI seconds.
3. Write one camera path: locked, dolly in, truck left, tilt up.
4. Add aspect ratio clause: 16:9, 9:16, 2.39:1 scope.
5. Add one lens family: 35mm spherical, anamorphic scope, 85mm portrait.
6. Add one grain tier: fine, medium, or heavy.
7. Read aloud. Delete duplicate cinematic adjectives.

Merge steps 4–6 into one trailing sentence when the host prefers prose over labels. Keep the order stable so edits stay predictable across hosts.

Image-to-video vs text-to-video

Text-to-video needs the full stack in prompt text because every pixel is invented. Aspect, lens, and grain all belong in the paste.

Image-to-video inherits composition and lens feel from the upload. Your text should lead with preserve language: "Maintain reference composition and lighting." Add motion and camera next. Add grain only if you want to override the still's clean digital look. Add aspect reminder if the host tends to re-crop: "Keep 16:9 framing from reference."

Crop the hero still to final aspect before upload. PromptMake /image can draft still prompts for Midjourney, FLUX, DALL·E, Stable Diffusion, or Leonardo when you need a widescreen hero before motion. Soft path: https://promptmake.net/image. Recreate Exactly holds layout. Adjust Lighting fixes exposure when grain would hide shadow detail.

Model notes for Kling, Runway, Luma, and Pika (2026)

Public model names rotate. Treat the notes below as an August 2026 snapshot. Confirm pickers inside each host before you paste forum prompts from last quarter.

Shared rule: SMCD or subject-motion-camera-duration first, then aspect, lens, grain. All four hosts punish contradictory look stacks. Scope ratio plus vertical lane plus fisheye plus heavy grain in one pass yields neither scope nor vertical.

Free tiers on these hosts favor short clips. Look keywords cost the same credits as plain prompts. Spend early tests on five-second clips with fine grain before you chase ten-second anamorphic hero shots.

Kling and Runway

Kling AI reads clause-separated stacks well. Put aspect in its own short phrase: "16:9 horizontal." Follow with lens: "35mm spherical, shallow depth." End with grain: "subtle 35mm film grain." Kling image-to-video preserves upload identity better when grain stays fine and camera stays to one move.

Runway Gen-4 targets film vocabulary. Anamorphic, scope ratio, and lens flare keywords align with Gen-4 training language in mid-2026 docs. Runway tolerates slightly richer lens stacks than Kling if camera motion stays single-path. Gen-4 longer lanes (eight to ten seconds) can hold a slow push-in with medium grain where shorter hosts flicker.

Both Kling and Runway expose aspect in UI. Repeat ratio in text anyway on first tests. Log which host ignores text ratio so you know when upload crop alone must carry composition.

Luma and Pika

Luma Dream Machine favors readable sentences on five-second clips. Lead with subject and motion in plain prose. Trail aspect and lens: "...in a 16:9 frame, 50mm shallow depth, light film grain." Luma image-to-video works best with subtle look overrides. Heavy anamorphic flare on upload often fights the reference pixels.

Pika focuses on short social clips as of August 2026. Keep cinematic keyword stacks short: ratio, one lens tag, optional fine grain. Pika duration limits reward one motion beat. Save 2.39:1 scope language for hosts with explicit scope export unless Pika's picker shows ultrawide.

Luma text-to-video accepts natural ordering. Pika accepts compact tags. When you port the same look stack across hosts, rewrite grain density down for Luma portrait uploads and shorten lens lists for Pika.

Common cinematic video keyword mistakes

Mistake 1: Still-image cinematic paste with no motion verbs. Grain and lens alone do not animate a frame.

Mistake 2: Ratio in prompt fights ratio in UI. Align both or crop upload.

Mistake 3: Three lens families in one line. Pick one focal length per pass.

Mistake 4: Heavy grain on fast action or label reads. Texture eats detail that motion needs.

Mistake 5: Anamorphic flare on every highlight. One motivated flare source reads cinema. Five flares read broken lens.

Mistake 6: Scope aspect on vertical export. Host crops or invents bars.

Mistake 7: Look keywords before subject anchor. Fix nouns and motion before you tune grain.

Mistake 8: Ignoring upload noise. Source still grain plus prompt grain doubles flicker on image-to-video.

When PromptMake /image fits cinematic video prep

Video hosts animate pixels. They do not fix a badly composed still. PromptMake /image helps you lock a hero frame in widescreen or vertical before you open Kling or Luma.

Use /image when you need a 16:9 or 2.39:1 still from a phone photo, when your upload keeps drifting because the source was square, or when you want Midjourney or FLUX dialect from a reference without hand-translating flags.

Workflow: build or upload a still, copy formatted prompt text, iterate until thumbnail-clear, export PNG cropped to target aspect, upload to image-to-video, paste motion and camera lines plus light grain if needed.

Goal modes map to video prep. Recreate Exactly holds layout for uploads. Create Variation explores alternate heroes with the same ratio. Adjust Lighting fixes blown highlights before you add grain in the video host.

Guest accounts receive about three /image runs per day. Free registration raises image quota separately from /text. PromptMake does not render video. It shortens the still step so cinematic look keywords attach to a frame worth animating.

FAQ

What are cinematic video prompts?

Cinematic video prompts describe short clips with subject, motion, camera, and look cues that survive temporal change. Look cues include aspect ratio language, film grain density, and lens or focal length tags. They differ from still-image prompts because video hosts need timed motion and one camera path, not a frozen lighting study alone.

Which aspect ratio keywords work for Kling and Runway?

Name the ratio explicitly: 16:9 horizontal, 9:16 vertical, 21:9 ultrawide, or 2.39:1 anamorphic scope. Match the same ratio in the host UI and crop image-to-video uploads to that lane. Pair ratio with composition hints like lower-third subject placement on vertical clips.

How much film grain should I put in a video prompt?

Start with one grain tier: fine for portraits and products, medium for documentary street, heavy only for deliberate retro or night mood. Add a second grain adjective after a clean render, not before. Reduce grain when motion is fast or when text on products must stay readable.

What lens keywords do Luma and Pika read best?

Luma reads focal length numbers plus shallow depth in a trailing sentence on five-second clips: 35mm spherical, 50mm medium, 85mm portrait, anamorphic with oval bokeh. Pika reads shorter tags: wide cinematic lens, anamorphic look, shallow depth. One lens family per clip on both hosts.

How is this different from cinematic lighting keywords for stills?

Still-image lighting guides cover key, fill, rim, Kelvin, and motivated practicals for one frame. This guide covers aspect, grain, and lens for motion hosts. Use the still lighting article when you generate heroes in Midjourney or FLUX. Use this article when you paste into Kling, Runway, Luma, or Pika.

Can I use cinematic video keywords on image-to-video uploads?

Yes, with restraint. Upload carries composition and lens feel. Lead with preserve language, add motion and camera, then add fine grain or anamorphic cues only if you want to override the still. Crop to target aspect before upload. Avoid heavy grain stacks on identity-sensitive portraits.

How do I start with cinematic video prompts for free?

Pick one host you can access, set a five-second duration, and paste a full stack: subject, motion, one camera move, 16:9 or 9:16 ratio, one lens tag, fine grain. Generate once. Fix motion before look. For hero stills first, use PromptMake /image at https://promptmake.net/image (about three guest runs per day, more when registered).