Characters are where AI video gets hard — the same face has to survive from shot to shot, mouths have to match words, and a glance has to carry a story beat. Runway Gen-4.5 handles all three better than most, but only if you prompt for it deliberately. Below are 20 copy-paste prompts across five story jobs, each built with concrete camera, lighting, and expression direction.
Three features do the heavy lifting. Character consistency: feed the same reference image into Image-to-Video and repeat the character's fixed features and wardrobe verbatim in every prompt. Act-Two is performance capture — record a driving video and it maps your face, hands, and full-body gestures onto the character. Lip Sync animates mouths to audio up to 45 seconds, from text-to-speech or an uploaded file, across up to 4 faces in one shot. Prompts below flag which feature to pair with each shot.
Consistent recurring characters
Consistency comes from repetition, not luck. Lock a reference image of your character, drive every shot with it in Image-to-Video, and copy the same appearance sentence into each prompt — same hair, same wardrobe, same defining features. When you use Image-to-Video, your text should describe motion only, since the reference already sets the look.
1. The recurring lead — establishing portrait
A woman in her early thirties, shoulder-length auburn hair, small scar above her left eyebrow, wearing a faded olive field jacket, stands at a rain-streaked window; she slowly turns her head toward camera and holds a steady, guarded gaze. Interior apartment, overcast daylight, soft diffused window light with cool shadows. Slow push-in on her face. Muted cinematic color, 16:9, 5s.Best for: the first "hero" shot — screenshot this frame as your reference image for every later clip.
2. Same character, new location
Same woman, early thirties, shoulder-length auburn hair, scar above her left eyebrow, faded olive field jacket, now walking through a crowded night market; she scans the stalls warily and pulls the jacket tighter. Neon signage and warm practical lights, wet pavement reflections. Tracking shot following her from the side at shoulder height. Handheld energy, 16:9, 10s.Best for: keeping identity locked across a scene change — start from the same reference image, keep the appearance wording identical.
3. Consistent villain across shots
A tall man, shaved head, deep-set eyes, thin silver ring on his index finger, wearing a charcoal three-piece suit, sits at the head of a long boardroom table and taps the ring slowly against the wood; a faint, humorless smile forms. Low-key chiaroscuro lighting, single hard rim light from behind. Slow orbit around him from left to right. Desaturated, 21:9 cinematic, 10s.Why it works: a single memorable prop (the ring) plus fixed features give Gen-4.5 clear anchors to reproduce.
4. Recurring child character at play
A seven-year-old boy with curly dark hair, gap-toothed grin, wearing a striped yellow shirt and red sneakers, runs across a sunlit backyard chasing a paper kite; he laughs and glances back over his shoulder. Golden hour, warm backlight and long grass shadows. Low-angle tracking shot moving with him. Nostalgic warm grade, 16:9, 5s.Best for: family or flashback threads — reuse the striped shirt and kite as continuity anchors.
Dialogue with Act-Two & Lip Sync
For talking characters, generate the shot first, then add the voice. Lip Sync syncs the mouth to TTS or uploaded audio (up to 45s, up to 4 faces); Act-Two maps a full performance — expression, hands, gestures — from a video you record. Quoted lines in these prompts are delivery cues for pacing and tone, not on-screen captions.
5. Monologue to camera (Act-Two)
A weathered fisherman in his sixties, white stubble, thick knit sweater, sits in a dim harbor pub and speaks directly to camera with quiet gravity, occasionally gesturing with an open hand as if recounting a memory. Warm practical lamplight, deep shadows behind him. Locked-off medium close-up, eye level. Grainy film look, 16:9, 10s.Best for: driving with Act-Two — record yourself delivering the monologue and map the performance, including the hand gestures, onto the fisherman.
6. Voicemail confession (Lip Sync)
A young man in a parked car at night, phone pressed to his ear, speaks in a low, unsteady voice — "I should have told you the truth sooner" — his eyes glassy, jaw tight. Dashboard glow and passing headlight streaks across his face. Static close-up from the passenger seat. Cool blue night grade, 16:9, 10s.Best for: generate the shot, then run Lip Sync with a recorded voice line so the mouth matches the confession exactly.
7. News anchor read (Lip Sync TTS)
A polished news anchor in a navy blazer sits at a studio desk facing camera, calm and authoritative, subtle head movements as she delivers a report. Bright high-key studio lighting, soft key and fill, blurred newsroom behind her. Locked-off medium shot, eye level. Clean broadcast look, 16:9, 10s.Best for: Lip Sync with text-to-speech — type the script, pick a voice, and let it drive the mouth for a hands-off talking head.
8. Gestural storyteller (Act-Two full body)
A lively street performer in a patched velvet coat stands on a cobblestone square and tells an animated story, arms sweeping wide, leaning in and back with the beats. Overcast diffused daylight, damp stone underfoot. Medium-wide shot, slight low angle, locked off. Painterly cinematic grade, 16:9, 10s.Why it works: a wide framing gives Act-Two room to map full-body gestures, not just the face.
Emotional close-ups & reactions
Reaction shots are about micro-expression and light. Keep the camera close and mostly still, name the exact emotional transition you want (from X to Y), and let Gen-4.5's motion model carry the shift. For continuity, drive these from your character's reference image.
9. Single tear close-up
Extreme close-up of a woman's face, eyes brimming; a single tear breaks and slides down her cheek as her expression holds between grief and acceptance, lips pressed thin. Soft window light from one side, gentle falloff into shadow. Locked-off macro shot, shallow focus on the eyes. Desaturated, delicate film grain, 16:9, 5s.Best for: the emotional peak of a scene — pair with a quiet score in the edit.
10. Slow-dawning smile
Close-up of a man reading a letter off-camera; his face shifts slowly from tense worry to disbelief to a wide, relieved smile as his eyes glisten. Warm golden hour light through a nearby window, soft rim on his hair. Very slow push-in on his face. Warm natural grade, 16:9, 10s.Why it works: naming the three-stage transition gives the model a clear emotional arc to animate.
11. Suppressed rage
Tight close-up of a man's face as he listens to something off-screen; his jaw clenches, a muscle twitches at his temple, and his nostrils flare while he forces his expression to stay flat. Hard low-key side light, deep shadow across half his face. Static shot, eye level, shallow focus. Cold desaturated grade, 16:9, 5s.Best for: tension beats where the character is holding something back.
12. Relief after bad news passes
Close-up of a woman standing in a hospital corridor; she exhales slowly, shoulders dropping, eyes closing for a moment before a fragile, grateful half-smile appears. Cool fluorescent overhead light softened by an out-of-focus window behind her. Slow handheld drift toward her. Naturalistic grade, 16:9, 10s.Why it works: the breath and shoulder drop give the model a physical action to hang the emotion on.
Two-person scenes
Two characters mean two identities to hold and a clear spatial relationship. Establish who is where, use over-the-shoulder framing to keep both in play, and if they speak, remember Lip Sync handles up to 4 faces in one shot.
13. Tense diner confrontation
Two people face each other across a chrome diner booth at night — a woman with cropped blonde hair, arms folded, and a nervous man in a rumpled shirt who avoids her eyes. She stares him down; he fidgets with a coffee cup. Warm neon and practical booth lights, deep window darkness outside. Slow push-in over her shoulder onto his face. Moody film-noir grade, 16:9, 10s.Best for: dialogue setups — over-the-shoulder keeps both characters anchored in frame.
14. Reunion embrace
Two friends spot each other across a busy train platform, break into a run, and collide into a tight, laughing embrace, one lifting the other slightly off the ground. Late afternoon sun raking through the station roof, dust and steam in the light. Camera cranes up and orbits gently around them. Warm nostalgic grade, 21:9, 10s.Why it works: one clear action plus one clear camera move keeps a busy shot readable.
15. Whispered secret over-the-shoulder
Two coworkers at the edge of a crowded office party; one leans in close to murmur something into the other's ear, whose eyes widen and flick sideways with alarm. Warm string lights and blurred partygoers behind them. Static over-the-shoulder close two-shot, shallow focus. Soft warm grade, 16:9, 5s.Best for: plot reveals — the widening eyes carry the beat without any spoken line.
16. Negotiation across a desk
Two executives sit opposite each other across a wide glass desk — an older woman leaning back with a cool, appraising look, and a younger man leaning forward, jaw set, pressing his case with a raised hand. Cool office daylight, hard reflections on the glass. Slow lateral tracking move revealing both profiles. Crisp corporate grade, 16:9, 10s.Best for: layering Lip Sync on both faces afterward for a full negotiation exchange.
Action heroes & story beats
Action beats reward one decisive movement plus a matching camera move. Keep the character's look fixed for continuity, and let physics and motion — a Gen-4.5 strength — sell the moment. Save these frames as references so your hero stays the same hero across the sequence.
17. Hero walks from the wreckage
A battle-worn woman in scorched tactical gear, short dark hair matted with dust, walks slowly toward camera away from a burning vehicle behind her, unflinching as embers drift past. Orange firelight and heavy volumetric smoke, hard backlit rim. Low-angle slow dolly back as she advances. Gritty high-contrast grade, 21:9, 10s.Best for: a signature "hero reveal" — the low angle makes her dominate the frame.
18. The reluctant draw
A grizzled gunslinger in a dust-caked duster stands in an empty desert street, hand hovering over his holster, eyes narrowing as sweat traces his cheek; his fingers flex once, hesitating. Harsh midday sun, hard shadows, shimmering heat haze. Slow push-in on his eyes, then a fast whip pan toward his hand. Bleached warm grade, 21:9, 10s.Why it works: the whip pan is timed to the beat, and one clear action keeps the short clip clean.
19. Rooftop leap decision
A courier in a weathered leather jacket stands at the edge of a rain-slicked rooftop, city lights sprawling below; she steadies her breath, coils, and launches into a running leap toward the next building. Neon-tinged night, wet reflections, cool volumetric haze. Camera cranes up and back as she jumps, holding her mid-air. Cinematic teal-orange grade, 16:9, 10s.Best for: a mid-sequence turning point — extend the clip to land the jump on the far roof.
20. Final stand rally
A young commander in dented armor turns to face a ragged line of soldiers behind her, raises her sword, and shouts a rallying cry; her expression hardens from fear to fierce resolve as the group surges forward. Stormy overcast light, cold rain, mud underfoot. Camera arcs from behind her shoulder around to a low hero angle. Desaturated epic grade, 21:9, 10s.Best for: the climactic beat — pair with Act-Two to drive the exact rallying performance from your own delivery.
Want the full library across every shot type? See the 35 best Runway prompts roundup, and grab reusable skeletons from the Runway prompt templates. To animate an existing character reference frame, the image-to-video prompts cover motion-only wording in depth.
Frequently Asked Questions
How do I keep a character consistent across Runway shots?
Use the same reference image in Image-to-Video for every shot, and re-state the character's fixed features and wardrobe verbatim in each text prompt. Gen-4.5 carries identity well when the anchor image and appearance wording stay identical between clips.
What does Act-Two do?
Act-Two is Runway's performance-capture feature. It maps the face, hands, and full-body gestures from a driving video you record onto your generated character, so the character performs your exact expressions and movements.
What are Lip Sync's limits?
Lip Sync animates mouths to audio up to 45 seconds long, driven by either text-to-speech or an uploaded audio file, and can sync up to 4 faces in a single shot.
Can Runway show my dialogue as on-screen text?
Not reliably. Gen-4.5 struggles with long or multi-line readable text, so treat quoted dialogue in a prompt as a performance and delivery cue, not as captions. Add real subtitles in an editor afterward.
Should I describe motion or appearance in Image-to-Video?
Describe motion only. The reference image already fixes appearance, wardrobe, lighting, and composition, so your text should cover the camera move and how the character moves, blinks, or emotes.
How long can a character-driven clip be?
Each Gen-4.5 generation runs 2–10 seconds. Use the 10s preset for dialogue and multi-beat performances, then extend clips or stitch them in an editor to build longer scenes.