How we made AI Video Editor with Magic Hour

Runbo Li
Runbo Li
·
· 15 min read
The Black Sports Car: original production reference 1 from AI Video Editor, source page 8.

Quick answer

AI Video Editor uses two character profiles and nine shot entries to plan a downtown automotive sequence with a deliberate wardrobe change. This breakdown includes all 12 original prompt texts and 15 reference images for Alan, Lisa, the passing sedan, black sports car, smoke compositions and final car interior. Use the references to check identity, screen direction and clothing continuity between shots.

AI Video Editor is a separate short film about a controlled city sequence. Alan’s dark overshirt changes to a leather jacket while his face, black T-shirt and street setting provide visual continuity. Lisa and the cars establish the relationships the camera needs to preserve.

The production board contains two character profiles and nine shot entries. One page has two distinct prompt fragments, so the library below retains twelve prompt texts rather than combining or inventing missing instructions.

Watch the final AI Video Editor film

The original prompt wording is retained; PDF line wrapping and repeated spaces have been cleaned. These are production directions and reference frames, not a controlled model benchmark. The boards do not establish generation cost, production time, attempt count or a guaranteed result.

How to use this production breakdown

Approve character references before composing a scene. Build a shot list that separates the input frame, action, camera movement and acceptance check for each shot.

Prepare reference stills with AI Image Generator. Use Image-to-Video when a motion prompt starts from an approved image. Keep the source images in the correct order: instructions that mention “the first image” or “the second image” depend on it.

The cinematic video prompt guide explains the reusable prompt structure. This article follows one specific film and preserves its original production material.

Nine prompts fit the native copyable cards. Three longer prompts are retained in full as quotations; no source wording has been shortened to fit the card limit.

Before you start: Alan and Lisa

Alan

What this entry specifies: Lock in Alan’s facial identity and original black T-shirt and overshirt.

Alan: original production reference 1 from AI Video Editor, source page 2.Alan: original production reference 2 from AI Video Editor, source page 2.

Original prompt

A photorealistic cinematic medium close-up of the same young adult man from the reference image, with naturally tousled, voluminous dark hair, subtle facial stubble, warm medium skin tone, expressive dark eyes, and a calm, confident half-smile. He wears a plain black crew-neck T-shirt under a dark navy-black casual overshirt with a visible collar.

Lisa

What this entry specifies: Establish Lisa’s facial identity and hair reference before composing the scenes.

Lisa: original production reference 1 from AI Video Editor, source page 3.Lisa: original production reference 2 from AI Video Editor, source page 3.

Original prompt

A photorealistic cinematic medium close-up of the young adult girl from the reference image, with wavy groomed smooth hair, voluminous tan brown red hair, fair skin, warm medium skin tone, expressive dark eyes, and a calm, confident half-smile.

Lisa’s source prompt uses both “fair skin” and “warm medium skin tone.” That wording is preserved. Approve the visual reference you intend to carry forward before generating the rest of your sequence.

Shot 1: Alan in the Financial District

What this entry specifies: Start with a restrained downtown close-up: eye contact, subtle breathing and natural daylight.

Alan in the Financial District: original production reference 1 from AI Video Editor, source page 4.

Original prompt

A photorealistic cinematic medium close-up of the same young adult man from the reference image, with naturally tousled, voluminous dark hair, subtle facial stubble, warm medium skin tone, expressive dark eyes, and a calm, confident half-smile. He wears a plain black crew-neck T-shirt under a dark navy-black casual overshirt with a visible collar. He stands on a busy downtown financial district street surrounded by tall modern glass-and-stone skyscrapers. Pedestrians move naturally in the distant background while cars pass along the road. The background is softly blurred with realistic depth separation and subtle circular bokeh from traffic lights and illuminated office windows. Action: He maintains relaxed eye contact with the camera, his head tilted slightly to one side. His expression is composed and quietly confident, with subtle natural breathing and minimal facial movement. A faint breeze gently moves a few strands of his hair. Camera: Sony FX6, 50mm cinema lens, eye-level medium close-up, shallow depth of field, focus locked on the eyes. Static camera with a barely perceptible handheld cinematic feel. Lighting and grade: Soft natural daylight filtering between skyscrapers, gentle highlights on his face, cool neutral shadows, realistic skin texture, restrained cinematic contrast, muted urban colors, natural highlights and shadows. Negative prompts: No face morphing, no hairstyle changes, no wardrobe changes, no artificial skin smoothing, no exaggerated expressions, no unnatural eye movement, no warped buildings, no distorted pedestrians, no excessive camera shake, no oversaturated colors.

Shot 2: Walking Toward the Camera (Frame 3)

What this entry specifies: Track backward with Alan’s walk while preserving identity, wardrobe and the street’s depth.

Walking Toward the Camera (Frame 3): original production reference 1 from AI Video Editor, source page 5.

Original prompt

A photorealistic cinematic wide frontal shot of the exact same young adult man from Frames 1 and 2, preserving his face, hairstyle, skin tone, black crew-neck T-shirt, and dark overshirt. He is centered on a broad downtown financial district street lined with tall modern skyscrapers, glass office towers, stone façades, black streetlamps, and green trees along the right-hand sidewalk. The scene takes place during bright late-afternoon daylight. The sky between the buildings is softly overexposed, creating a natural luminous backdrop. Glass façades reflect the daylight, while the street and trees retain realistic detail. A few pedestrians and vehicles move naturally in the background. Action: The man walks forward toward the camera with a relaxed, confident posture. His shoulders remain loose, his arms swing gently in opposition to his steps, and his head stays relatively steady. He briefly narrows his eyes against the daylight and maintains a subtle, composed expression. His movement feels spontaneous and human, with natural weight transfer and accurate foot contact against the pavement. Camera: 35mm cinema lens, wide frontal composition, camera positioned around chest height. The camera smoothly tracks backward at the same speed as the man's walking pace, keeping him centered and approximately the same size in the frame. Background parallax reveals the depth of the street while the subject remains visually stable. Natural motion blur only on moving background elements. Lighting and grade: Soft late-afternoon daylight, bright highlights between skyscrapers, neutral skin tones, slightly cool shadows, realistic contrast, restrained cinematic color grading, natural texture and subtle film grain. Composition: Widescreen 2.39:1 cinematic framing. Show the man from approximately mid-thigh upward, with the skyscrapers rising on both sides and a clear central vanishing point behind him. Negative prompts: No face morphing, no identity drift, no wardrobe changes, no running, no exaggerated walking, no foot sliding, no floating body movement, no distorted hands, no jittering face, no sudden zooms, no flickering windows, no warped buildings, no artificial background movement, no text or logos.

Use this complete prompt with Image-to-Video. Keep the matching reference images attached.

Shot 3: Blue-Hour Finger Gesture

What this entry specifies: Contrast deep-blue twilight with warm city lights around one controlled hand gesture.

Blue-Hour Finger Gesture: original production reference 1 from AI Video Editor, source page 6.

Original prompt

A cinematic wide shot featuring the exact same young adult man from Frame 1. Preserve his facial identity, tousled dark hair, black crew-neck T-shirt, and dark overshirt without alteration. He stands in the center foreground of a grand downtown street canyon, surrounded by towering skyscrapers with illuminated windows. It is blue hour, immediately after sunset. The open strip of sky between the buildings is rich deep blue, contrasting with warm amber streetlights, glowing storefronts, and golden reflections on the pavement. Vehicles pass behind him, producing realistic elongated white and red light trails while the man remains sharply defined. Action: The man faces the camera with a relaxed, self-assured expression. He raises one hand in front of his upper torso and performs one small, natural finger gesture, as if casually signaling or snapping his fingers. His wrist and fingers move with anatomically correct articulation. He then settles into a relaxed stance, keeping his gaze toward the camera. The gesture is subtle and confident, never theatrical. Camera: Wide 28mm anamorphic cinema lens, slightly low camera angle, medium-long distance, symmetrical urban composition with the man centered. A slow, subtle dolly-in creates depth while keeping the surrounding skyscrapers prominent. Natural motion blur on passing traffic, stable facial detail. Lighting and grade: Deep blue twilight ambience, warm tungsten street lamps, practical building lights, gentle warm edge light on his hair and shoulders, cinematic amber-and-blue color contrast, realistic film grain, rich blacks, natural skin tones. Composition: Widescreen 2.39:1 cinematic framing. Keep the man visible from approximately the waist up, with generous architectural space above and around him. Negative prompts: No identity changes, no extra fingers, no hand deformation, no exaggerated snapping, no floating limbs, no sudden camera jumps, no bending skyscrapers, no duplicated cars, no artificial light trails passing through the character, no text, no logos, no unnatural facial animation.

Use this complete prompt with Image-to-Video. Keep the matching reference images attached.

Shot 4: The Passing Sedan

What this entry specifies: Establish Alan’s eyeline and the champagne-beige sedan’s position within the street.

The Passing Sedan: original production reference 1 from AI Video Editor, source page 7.

Original prompt

A photorealistic cinematic still of the same young adult man in the right foreground of a downtown street, shown in a three-quarter side profile. He has tousled dark hair, warm medium skin, subtle stubble, and wears a black crew-neck T-shirt beneath a dark navy-black casual overshirt. His head is turned toward the left side of the frame, his eyes focused on a passing vehicle, with a calm, observant expression. His shoulders remain oriented mostly forward while his face turns naturally toward the street. A champagne-beige older-model four-door sedan occupies the center-left middle ground, traveling along the road parallel to the sidewalk. The sedan has rounded bodywork, silver wheel covers, clear windows, and realistic reflections. Its front faces toward the left side of the image, and its wheels and body proportions are anatomically and mechanically accurate. Behind the car is a row of modern glass-and-stone office buildings, a transparent entrance canopy, street trees, grey pavement, and a few pedestrians in business-casual clothing. The man is sharply focused in the foreground while the sedan and distant pedestrians have moderate cinematic background blur. Eye-level camera, 50mm lens, side-oriented composition with the man occupying the right third and the beige sedan filling the left-center. Soft natural daylight, neutral urban color grading, realistic textures, subtle film grain, widescreen 2.39:1 composition with black letterbox bars. Maintain exact character identity and outfit continuity. No black sports car in the foreground, no text, no logos, no distorted wheels, no duplicated vehicles.

Shot 5: The Black Sports Car

What this entry specifies: Arrange Alan, the parked sports car and the woman in one readable composition.

The Black Sports Car: original production reference 1 from AI Video Editor, source page 8.

Original prompt

A photorealistic cinematic medium close-up of the same young adult man from the reference images, with naturally tousled dark hair, warm medium skin, subtle stubble, expressive dark eyes, and a relaxed, confident half-smile. He wears a plain black crew-neck T-shirt and a dark navy-black casual overshirt. His face and upper body occupy the right third of the composition, with his head slightly tilted toward the camera and his eyes looking directly at the viewer. A sleek, low-profile black exotic sports car is parked parallel to the curb in the middle background, positioned diagonally across the center-left of the frame. The car has glossy black bodywork, sculpted aerodynamic panels, dark wheels, a low roofline, and realistic reflections of the surrounding buildings. A young woman with long brown hair, an olive-green top, and a light patterned jacket stands on the far-left sidewalk, looking toward the car. The setting is a modern downtown business district with glass office façades, a transparent entrance canopy, pavement, street trees, and a few distant pedestrians. Eye-level camera, 50mm cinema lens, shallow depth of field, the man's face in sharp focus and the sports car recognizable in the background. Soft natural daylight, cool neutral color grading, realistic skin texture, cinematic contrast, widescreen 2.39:1 composition with black letterbox bars. Preserve the man's facial identity and wardrobe. No text, no logos, no deformed car body, no exaggerated smile.

Shot 6: Original Wardrobe Detail

What this entry specifies: Isolate the cotton overshirt and T-shirt as a clear wardrobe reference without a face.

Original Wardrobe Detail: original production reference 1 from AI Video Editor, source page 9.

Original prompt

A photorealistic cinematic still of the same young man, framed tightly from the base of his neck to his lower torso. He wears a plain black crew-neck T-shirt beneath an open dark navy-black cotton overshirt with a soft collar, visible front placket, subtle stitching, and a small chest pocket. The fabric is matte, slightly textured, and naturally folded around his shoulders and chest. His face is completely outside the frame. In the background on the right, the same young woman with long brown hair wears an olive-green top and a light patterned jacket, softly out of focus on a downtown sidewalk. Modern glass office buildings, a grey pavement, and part of a dark parked vehicle form the background. Centered torso composition, 85mm lens, shallow depth of field, soft overcast daylight, realistic fabric texture, muted urban colors, subtle cinematic grain, widescreen 2.39:1 framing with black letterbox bars. No leather jacket, no visible face, no wardrobe logos, no text, no distorted anatomy.

Shot 7: Leather Jacket Transformation

What this entry specifies: Specify the changed jacket through leather grain, collar, stitching and zipper detail.

Leather Jacket Transformation: original production reference 1 from AI Video Editor, source page 10.

Original prompt

A photorealistic cinematic still of a young man wearing a black leather biker-style jacket over a plain black crew-neck T-shirt, framed tightly from the base of his neck to his waist. The jacket features a broad folded collar, a partially visible silver metal zipper, realistic leather grain, natural creases, and subtle cool reflections. His face is entirely cropped out of the frame. The man occupies most of the foreground, with the leather jacket sharply detailed. In the softly blurred background on the right, a young woman with long brown hair wears an olive-green top, a light beige patterned jacket, and dark trousers. She stands on a modern downtown sidewalk beside glass office buildings. A dark vehicle is partially visible in the lower-left background. Eye-level chest-height composition, 85mm portrait lens, shallow depth of field, soft natural daylight, cool neutral color grading, realistic skin and fabric textures, cinematic contrast, widescreen 2.39:1 composition with black letterbox bars. No text, no logos, no visible face, no distorted clothing, no artificial leather texture.

Shot 8: The Smoke Transition

What this entry specifies: Preserve the two source references and their distinct scene descriptions.

The Smoke Transition: original production reference 1 from AI Video Editor, source page 11.The Smoke Transition: original production reference 2 from AI Video Editor, source page 11.

Reference A · left source column

Modern glass-and-steel office façade, grey stone pavement, dark entrance doors, realistic reflections of the building in the car roof. Eye-level medium-wide shot, 50mm cinema lens, foreground characters softly blurred and the newly arrived man relatively sharp. Soft overcast daylight, cool neutral urban color grade, natural skin textures, cinematic contrast, widescreen 2.39:1 aspect ratio with black letterbox bars.

Reference B · right source column

A photorealistic cinematic widescreen still on a modern downtown street outside a glass-and-steel office building. In the left foreground, a young man with tousled dark hair wears a black leather jacket over a black crew-neck T-shirt. He is seen from behind and in partial side profile, his head turned toward the center of the scene. In the right foreground, a young woman with long brown hair wears a beige-and-blue patterned jacket over an olive-green top. She looks toward the center with a curious, slightly surprised expression. A sleek black exotic sports car is parked between them and the building, its glossy roof and curved windows reflecting the surrounding glass architecture. Directly above and behind the car, near the building entrance, a large dense cloud of pale grey-white smoke hangs in the air, shaped like a human-sized silhouette that has just vanished. The cloud has soft, billowing edges, layered volumetric density, and realistic wisps dispersing into the surrounding air. The space within the cloud contains no visible human body.

Source note: Page 11 contains two separate text fragments: one mentions a newly arrived man; the other describes a vanished silhouette. Both are preserved without inventing an action between them.

Shot 9: Final Car-Interior Shot

What this entry specifies: Close on Alan in the sports car, carrying the same face, leather jacket and natural lighting into the interior.

Final Car-Interior Shot: original production reference 1 from AI Video Editor, source page 12.Final Car-Interior Shot: original production reference 2 from AI Video Editor, source page 12.

Original prompt

A photorealistic cinematic still of the same young man with tousled dark hair, warm medium skin, subtle stubble, and expressive dark eyes, seated in the driver's seat of a sleek black luxury sports car. He wears a black leather jacket with a broad folded collar and silver zipper over a plain black crew-neck T-shirt. His face is turned toward the camera, making direct eye contact with a relaxed, confident half-smile, as if addressing the viewer and delivering the final call to action of a premium commercial. His upper body is angled naturally toward the open driver's door. One arm extends toward the interior door panel, with his hand resting casually on the armrest. The car's dark leather upholstery, sculpted door panel, metallic trim, and curved roofline frame his face. Through the open doorway and side windows, a modern downtown street with glass-and-steel office buildings appears softly blurred in the background. Composition: Intimate medium close-up from just outside the open driver's door, camera positioned around the seated man's eye level. His face occupies the center-right of the frame, with the open door and interior trim creating strong diagonal leading lines from the left. His eyes and facial features are sharply focused while the urban background has a natural shallow depth of field. Lighting: Soft natural daylight enters through the open door, illuminating his face evenly with gentle highlights on his hair and realistic reflections across the black leather jacket. Subtle shadows inside the cabin create depth without obscuring his eyes. Visual style: Premium automotive commercial, cinematic realism, natural skin texture, authentic leather grain, realistic glass reflections, restrained cool-neutral color grading, subtle film grain, 50mm cinema lens aesthetic, widescreen 2.39:1 composition with black letterbox bars. Continuity: Preserve the exact facial identity, hairstyle, black leather jacket, and black T-shirt from the previous street scenes. The same black sports car is now the setting for the closing shot. Negative prompts: No face morphing, no exaggerated smile, no distorted fingers, no extra limbs, no warped dashboard, no artificial skin smoothing, no visible logos, no text, no subtitles, no steering wheel obstructing his face, no unnatural reflections.

Use this complete prompt with AI Image Generator. Keep the matching reference images attached.

Refine the approved sequence

Use the editor for targeted changes to an existing video. Describe the specific edit and check that the surrounding identity, composition and motion still work.

Explore AI Video Editor

Three decisions that structure this sequence

1. Separate identity from the wardrobe change

The original wardrobe close-up isolates the cotton overshirt and black T-shirt. The next reference specifies a leather biker jacket with its collar, stitching and zipper. Keep the deliberate change in its own brief, then review the face and T-shirt as fixed details when you carry the jacket into the car interior.

Use the character consistency guide to turn those fixed details into acceptance checks. An approved reference still helps define what to inspect; it does not prove that every generated shot will preserve it.

2. Establish screen direction before motion

The passing-sedan composition places Alan on the right, the car left-center and the vehicle’s front facing left. Preserve those relationships when choosing the reference frame. Decide the eyeline, vehicle position and travel direction before adding motion.

3. Match camera movement to the performance

The walking prompt tracks backward at Alan’s pace with a 35mm composition. The blue-hour gesture instead asks for a subtle dolly-in and a small gesture. Review the walk, hands, face and background as separate acceptance checks; a plausible single frame does not establish a smooth sequence.

What the source does and does not specify

The smoke page contains two separate still descriptions. One positions the newly arrived man; the other describes the vanished silhouette and explicitly excludes a visible human body. The source does not provide a complete animation instruction connecting them. Preserve the two compositions instead of inventing an additional transformation effect.

The walking entry is labeled Frame 3; no missing Frame 2 is supplied. Lisa’s prompt contains conflicting skin-tone wording, preserved above with a source note. Approve the intended visual reference before using it as a continuity anchor.

Before assembling the final edit, compare adjacent clips for identity, wardrobe, eyelines, car geometry, screen direction and lighting. Use the AI video preproduction checklist to establish inputs and acceptance criteria. For a commercial project, review the cinematic commercial workflow; this board does not establish distribution, music, trademark or likeness permissions.

Ready to create your own?

Start with the two character references, approve a street composition and test one deliberate change. Generate stills with AI Image Generator, animate approved frames with Image-to-Video, and use AI Video Editor for targeted edits to existing clips.

Explore Magic Hour when you are ready to build your own sequence.

A separate production breakdown: Good Agency

For an ensemble comedy with changing environments, read the Good Agency short-film breakdown. Its four characters, four locations and sixteen scene entries form a different production library.

Runbo Li
Runbo Li
CEO of Magic Hour
Runbo Li is the Co-founder and CEO of Magic Hour, where he builds AI video and image tools for content creation. He is a Y Combinator W24 founder and former Data Scientist at Meta, where he worked on 0-1 consumer social products in New Product Experimentation. He writes about AI video generation, AI image creation, creative workflows, and creator tools.
View author →

Continue Reading

T-Rex in the City Action: original production reference 1 from Good Agency, source page 23.

How we made Good Agency with Magic Hour

How we made Good Agency with Magic Hour
Cinematic AI Video Prompt Cookbook

25 cinematic AI video prompts for motion and composition

25 cinematic AI video prompts for motion and composition
Keep Characters Consistent in AI Video

How to keep characters consistent in AI video

How to keep characters consistent in AI video
A director's desk with hand-sketched storyboard squares, lens notes and shot order numbered only by visual ticks, camera and film strips

AI video shot list template: plan prompts, assets and review

AI video shot list template: plan prompts, assets and review
Comparison of top AI video generators including Magic Hour, Veo 3, Kling, Hailuo, Runway, and .

Best AI video generation workflow for cinematic commercials

Best AI video generation workflow for cinematic commercials
 (Free, No Drawing Skills Needed)

How to generate a storyboard with AI for free

How to generate a storyboard with AI for free