Runway Prompts

Act as a commercial director. I need a 10 second hero shot for [PRODUCT_NAME], a [PRODUCT_CATEGORY] from [BRAND_NAME], for use in [PLACEMENT] (homepage hero, ad, social). Write a Gen-4 text-to-video prompt under 1000 characters that describes the shot in this order: subject and action first, then environment, then lighting and mood, then camera move. Shot: the product [PRODUCT_ACTION] in a [SETTING] environment, [LIGHTING_STYLE] lighting, [COLOR_PALETTE] color palette, camera move [CAMERA_MOVE] (dolly in, slow orbit, or crane up), shot on [LENS_LOOK] look, aspect ratio [ASPECT_RATIO] (16:9 landscape or 9:16 vertical). Add generation settings: use a fixed seed [SEED] for reproducibility, generate [NUMBER_OF_VARIANTS] variants, and pick the take with the cleanest product geometry. Output format: the exact prompt text to paste into Runway, followed by the settings list.

Act as a motion designer. I have a still image of [SUBJECT_DESCRIPTION] in [STYLE] style that I want to bring to life with Runway Gen-4 image-to-video. Write the image-to-video prompt: describe ONLY the motion, not the still itself (the model already sees the image), specifying [MOTION_1] for the main subject, [MOTION_2] for a background element, and [ATMOSPHERE_MOTION] (dust, mist, light flicker). Set duration to [DURATION] seconds, aspect ratio matching the source [ASPECT_RATIO], and a fixed seed [SEED]. Include the rule: if the output distorts [CRITICAL_DETAIL] (faces, hands, product logo), regenerate with the motion described more simply before touching any other setting. Output format: the motion-only prompt text plus the settings and the distortion recovery rule.

Act as a performance director using Runway Act-One. I am directing a scene where my AI character [CHARACTER_NAME], a [CHARACTER_DESCRIPTION], must deliver the emotional beat [EMOTION_BEAT] (for example delivering bad news, a joyful reveal) in a [SETTING] environment. Plan the Act-One shoot: record a [DURATION] second reference performance video of [PERFORMER_DESCRIPTION] doing the full performance, framing [FRAMING] (head and shoulders, waist up, or full body), in [LIGHTING] lighting so facial nuance reads clearly. Upload the character reference image [CHARACTER_REF_DESCRIPTION], then map the performance onto the character, watching for [NUANCE_1] (weight shift), [NUANCE_2] (breathing), and [NUANCE_3] (hand gestures). If the transfer looks stiff, re-record the reference with [PERFORMANCE_ADJUSTMENT] (bigger, smaller, or slower). Output format: shoot plan, mapping checklist, and the re-record decision rule.

Act as a VFX editor using Runway Aleph. I have an existing [DURATION] second clip of [ORIGINAL_CLIP_DESCRIPTION] shot in [SETTING], and I need to change exactly one thing: [EDIT_VERB_AND_CHANGE] (for example replace the daytime sky with a stormy sunset, remove the extra person on the left, or change the car color from red to black). Write the Aleph edit instruction using the verb-first format: start with the verb ([ADD], [REMOVE], [REPLACE], [RELIGHT], or [RESTYLE]), then describe ONLY the change, leaving everything else untouched. Specify what must stay identical: [ELEMENTS_TO_PRESERVE] (camera move, actor performance, background). Generate [NUMBER_OF_VARIANTS] variants, review on a [REVIEW_METHOD] (frame-by-frame or side-by-side), and accept only the variant where [ACCEPTANCE_CRITERIA]. Output format: the exact Aleph instruction text plus the review checklist.

Act as a cinematographer in Runway. I need a [DURATION] second clip of [PRODUCT_NAME] rotating on a [SURFACE] surface against a [BACKGROUND] background for [USE_CASE] (ad, PDP video, social). Configure the shot: Camera Control set to [CAMERA_MOVE] (slow orbit at [ORBIT_SPEED] or dolly in), and Motion Brush painted on [BRUSH_REGION_1] (the product itself, [MOTION_DIRECTION_1] rotation) and [BRUSH_REGION_2] (a background element like steam or fabric, [MOTION_DIRECTION_2] drift), leaving [STATIC_REGION] completely still. Set ambient motion intensity to [INTENSITY] so the product stays sharp. Generate [NUMBER_OF_VARIANTS] variants and pick the one where the product label [LABEL_DETAIL] stays legible through the full move. Output format: camera settings, motion brush region map with directions, and the selection rule.

Act as a dialogue editor using Runway lip sync. I have a [DURATION] second clip of [CHARACTER_DESCRIPTION] who must speak the line '[DIALOGUE_LINE]' in [LANGUAGE] with [EMOTION] delivery. The audio file is [AUDIO_SOURCE] (ElevenLabs voiceover, recorded VO, or on-set audio). Plan the lip sync pass: trim the audio to start [LEAD_IN] ms before the mouth opens, run lip sync on the clip, and check [CHECK_1] (mouth shapes match plosives like b, p, m), [CHECK_2] (no drift by the end of the line), and [CHECK_3] (head motion still looks natural). If drift appears after [DRIFT_THRESHOLD] seconds, split the line into [SEGMENT_COUNT] segments and sync each separately. Output format: the prep steps, the sync checklist, and the drift recovery procedure.

Act as a cleanup artist using Runway inpainting. My [DURATION] second clip of [SCENE_DESCRIPTION] has an unwanted [OBJECT_TO_REMOVE] (boom mic, extra person, logo, trash) visible during [TIME_RANGE]. Plan the inpainting job: mask the object with a [MASK_STYLE] mask ([TIGHT_OR_LOOSE] around edges, feathered by [FEATHER] px), track the mask across the clip, and inpaint with the fill instruction '[FILL_DESCRIPTION]' (what should appear instead, e.g. clean wall, empty sky). Review the result for [ARTIFACT_1] (flicker), [ARTIFACT_2] (smearing on edges), and [ARTIFACT_3] (lighting mismatch with the plate). If artifacts persist, shrink the mask and run [NUMBER_OF_PASSES] passes. Output format: masking instructions, fill prompt, and the artifact review checklist.

Act as a finishing editor using Runway outpainting (Expand Video). I have a [SOURCE_ASPECT] clip of [SCENE_DESCRIPTION] that must be delivered in [TARGET_ASPECT] (for example 16:9 to 9:16 for [PLATFORM]). Plan the expansion: outpaint [DIRECTION_1] and [DIRECTION_2] (top and bottom, or left and right), describe the extension content as '[EXTENSION_DESCRIPTION]' (what plausibly continues the scene, matching [LIGHTING] and [COLOR_GRADE]), and keep the original frame untouched in the center. Generate [NUMBER_OF_VARIANTS] variants, check [CHECK_1] (no duplicated subjects in the extended area), [CHECK_2] (grain and sharpness match the original), and [CHECK_3] (the seam is invisible at [SEAM_CHECK_METHOD]). Output format: extension description, generation settings, and the seam QA checklist.

Act as a delivery supervisor. My final edit of [PROJECT_NAME], a [PROJECT_TYPE] for [CLIENT_OR_CHANNEL], is [SOURCE_RESOLUTION] and must deliver at [TARGET_RESOLUTION] ([DELIVERY_SPEC], e.g. 4K master, 1080p social). Plan the Runway upscaling pass: upscale in [SEGMENT_LENGTH] second segments to catch artifacts early, then review each segment for [ARTIFACT_1] (over-sharpened edges), [ARTIFACT_2] (face warping), and [ARTIFACT_3] (temporal flicker on fine detail like [FINE_DETAIL]). Set a reject threshold: any segment failing [REJECT_CRITERIA] gets re-upscaled with [ADJUSTMENT] (lower enhancement strength). After all segments pass, reassemble in [EDIT_TOOL] and do a final full watch at [WATCH_SPEED]. Output format: the segment plan, artifact checklist with reject rules, and the reassembly steps.

Act as a film director planning a [NUMBER_OF_SHOTS] shot brand film for [BRAND_NAME] about [CAMPAIGN_THEME], targeting [TARGET_AUDIENCE]. Use Runway References to lock consistency: ingest [CHARACTER_REF] (hero character), [LOCATION_REF] (key location), and [OBJECT_REF] (hero product) as reference assets. For each shot, write a Gen-4 prompt under 1000 characters that names the reference explicitly (e.g. 'the hero character from the reference, now [SHOT_ACTION]'), specifies the camera move [CAMERA_MOVE_PER_SHOT], and the lighting [LIGHTING_PER_SHOT], while keeping [CONSISTENT_ELEMENT] identical across all shots. Plan the edit order [SHOT_ORDER], transitions [TRANSITION_STYLE], and a consistency QA pass comparing [QA_COMPARISON] (character face, product label, location details) across shots. Output format: reference asset list, per-shot prompts, edit plan, and QA checklist.