AI Prompts for Runway Prompts for Cinematic B-Roll

My project: [DESCRIBE: YOUTUBE VIDEO / DOCUMENTARY SEGMENT / CLIENT FILM] about [TOPIC], with A-roll of [WHAT YOU HAVE: TALKING HEAD, SCREEN RECORDING, INTERVIEWS]. Build the B-roll coverage plan: the cutaway moments in my script that need visual support ([PASTE SCRIPT OR OUTLINE]), the shot type for each (establishing, detail, atmosphere, metaphor), duration needed, and the priority order if I can only generate half of it.

For my B-roll list ([PASTE SHOTS NEEDED]), triage each shot: generate in Runway (specific compositions, impossible or expensive shots, exact-match atmosphere), use stock (generic humans doing generic things, real locations that must be real), or shoot myself ([MY GEAR]). Decision criteria: specificity needed, human close-up realism, cost, and rights. Output the sourced shot list.

Define the visual language for my project’s B-roll so every generated shot belongs: reference mood ([FILMS / STYLES I WANT]), the lens feel (wide and immersive versus long-lens compression), light character (natural, moody, clinical), movement rule (locked off, drifting, handheld energy), color world, and time of day bias. Output as the style block I append to every Runway prompt.

My video needs establishing shots of: [LOCATIONS / SETTINGS, REAL OR CONCEPTUAL]. For each, design the cinematic establishing shot: composition, camera move (slow push, aerial drift, static with internal motion), light and weather for the mood ([MOOD PER LOCATION]), and the Runway prompt in full cinematic language. Establishing shots are where AI B-roll shines: no location fees, exact mood control.

My script discusses abstract ideas: [LIST CONCEPTS: GROWTH, RISK, CONNECTION, TIME PRESSURE]. Design a metaphor shot for each: the concrete image that carries the concept without being a cliche (no chess boards, no lightbulbs), the motion that makes it cinematic, and the Runway prompt. Also flag which cliches to avoid per concept, since AI defaults to them.

Write the Runway prompt for this shot: [DESCRIBE]. Build it in layers: subject and action (specific, singular), environment with depth cues (foreground element, midground subject, background falloff), light (source direction, quality, color temperature), lens and camera (focal feel, movement, speed), and grade (filmic, muted, high contrast). Then strip any word doing no work. Give me the final prompt plus the one-variable variants for the likely miss.

My generations look artificial and I think it is the light. Rewrite my prompts ([PASTE 2-3]) with realistic light logic: one motivated source per scene (window, sun angle, practical lamp), shadow behavior that matches, atmospheric interaction (haze catching light, bounce fill), and time-of-day consistency. Explain the light logic per rewrite so I internalize it.

AI video over-moves. Define my camera discipline for B-roll: when static with internal motion beats camera movement (dialogue support, contemplative beats), the two moves that always look professional (slow push-in, lateral drift), the speed rule (half as fast as feels right), and rewrite my shot list prompts ([PASTE]) applying the discipline. Restraint reads as intention; intention reads as cinema.

My B-roll looks too clean, too digital. Add the imperfection layer to my prompts: filmic texture terms (grain, halation, slight vignette), atmospheric particles (dust, mist, rain on glass), lens behavior (subtle flare, focus breathing), and natural imperfection in scenes (weathered surfaces, uneven light). Rewrite these prompts with the texture layer: [PASTE]. Perfection is the tell; imperfection sells real.

This generated shot must cut against my real A-roll footage: [DESCRIBE A-ROLL: SETTING, LIGHT, GRADE, LENS FEEL]. Write the Runway prompt matched to it: same light temperature and direction, similar contrast and color world, compatible lens compression, and motion energy that flows from the A-roll pace. Then list the edit-stage adjustments (grade tweaks) to close the remaining gap.

Design my B-roll library structure: categories that match how editors search ([SUGGEST: BY MOOD, BY SETTING, BY CONCEPT, BY MOTION TYPE]), the file naming convention (setting_subject_move_mood_v1), the metadata worth logging per clip (prompt used, style block version, project origin), and the tool ([OPTIONS: FOLDERS / DAM / SPREADSHEET INDEX]) fitting my scale: [CLIPS EXPECTED].

Plan a batch session generating evergreen B-roll I will reuse constantly: for my content about [TOPICS], the 15 most reusable shots (atmospheres, settings, textures, transitions) ranked by expected reuse, the Runway prompt for each in my visual language ([PASTE STYLE BLOCK]), and the session workflow: generate, cull to keepers, name, file. One afternoon, permanent asset.

Analyze my recent projects for B-roll gaps: here are the moments I struggled to cover or settled for weak stock ([DESCRIBE OR LIST]). Cluster the gaps into shot families, design the generation list that fills each family with 2-3 options, and flag the gaps AI still cannot fill well (real people close-up, brand-specific locations) with the sourcing alternative for each.

Schedule my library refresh: the seasonal shots worth generating ahead ([UPCOMING: SEASONS, HOLIDAYS, ANNUAL MOMENTS IN MY NICHE]), the visual trends in [MY CONTENT SPACE] worth testing ([DESCRIBE WHAT YOU SEE WINNING]), and the aging rule: which library clips look dated after [TIME] and need regeneration with current model quality. Quarterly calendar, one session each.

Set up my usage governance for generated B-roll: the license terms of my Runway plan for commercial client work (what to verify), the per-client usage log (which generated clips shipped in which deliverables), the disclosure decision for client projects (my policy and contract line), and the one rule for client work: never generate B-roll depicting identifiable real people, places, or brands without checking rights first.

Map the B-roll cuts for my edit: [PASTE SCRIPT / A-ROLL TRANSCRIPT WITH TIMESTAMPS]. For each cutaway: the motivation (illustrating the point, covering an edit, emotional beat, pacing relief), the clip from my library or generation list that serves it, the cut-in and cut-out logic (on the action, on the beat), and the density check: [STYLE] content wants B-roll [PERCENTAGE] of runtime, not wall-to-wall.

My B-roll cuts feel arbitrary. Give me the rhythm system: shot duration matched to content energy (fast cuts for momentum sections, long holds for weight), the music-grid option (cutting on bars) versus content-driven cuts, minimum shot duration before it reads as a flash, and the variety rule: consecutive B-roll shots must change size or angle. Apply it to my sequence: [DESCRIBE CURRENT CUT].

My edit mixes Runway B-roll, [OTHER SOURCES: STOCK, MY FOOTAGE]. Plan the grade that unifies them: the common target (contrast curve, color temperature, saturation ceiling), the per-source correction needed (generated footage tends toward [TENDENCIES: OVER-SATURATION, PLASTIC HIGHLIGHTS]), the film emulation or LUT approach for cohesion, and the QA: skim the timeline at speed, any shot that announces its source needs work.

Add sound to my B-roll moments: [LIST THE B-ROLL SEQUENCES IN MY EDIT]. For each: the ambient bed that sells the location (generated footage is silent; sound makes it real), the spot effects worth adding (footsteps, wind, room tone shifts), and the mix level under my A-roll voice. Silent B-roll is the second biggest AI tell after bad hands.

Review my B-roll usage across this finished edit: [DESCRIBE THE VIDEO AND WHERE B-ROLL LANDS]. Audit: does every cutaway earn its place or is any decorative filler, do any generated shots break immersion (artifacts, mismatched light), is the best shot placed at the moment of highest impact, and what does this edit teach my library: shots to generate before the next project.