FORMAT: A multi-camera edit of one continuous take. The number of shots and the camera order are NOT chosen freely — they are looked up in the ROUTER below according to the measured duration of @video_1 and the movement mode detected in STEP 2. Camera mov

Видео-обложка

FORMAT: A multi-camera edit of one continuous take. The number of shots and the camera order are NOT chosen freely — they are looked up in the ROUTER below according to the measured duration of @video_1 and the movement mode detected in STEP 2. Camera movement is permitted ONLY where this prompt explicitly grants it. Do not use fewer shots than the router specifies. TASK: @video_1 is the source take. Do not re-perform it and do not re-time it. Re-observe the exact same performance from a set of new camera positions, as if a full camera crew had covered the original take simultaneously, and cut between them. The subject's performance, the environment, the wardrobe and the lighting are fixed. Only the point of view changes. STEP 1 — MEASURE THE SOURCE: Read the exact duration of @video_1 in seconds. Read the speech rhythm: sentence boundaries, breaths, pauses, emphasis beats. Read the subject, wardrobe, hair, skin, props, background geometry, lens character, grain and colour science and keep them pixel-consistent in every shot. STEP 2 — DETECT THE MOVEMENT MODE. This decision comes before everything else. Look at whether the subject physically travels through space in @video_1. MODE STATIC — the subject is seated or standing on one spot. His body stays at the same coordinates for the whole take. Only his head, hands and torso move. MODE MOVING — the subject walks, strolls, paces or otherwise travels through space while speaking. His position in the world changes over the duration of the take. If in doubt, choose MODE STATIC. A wrong movement decision is the single most damaging error in this task. STEP 3 — CLASSIFY THE LOCATION. TYPE A — office, business centre, meeting room or lobby with a window wall or large windows. TYPE B — cafe, restaurant, bar or coworking space with a street-facing window. TYPE C — open air: street, pavement, terrace, rooftop, courtyard, park, embankment. TYPE D — enclosed room with no windows and no view outside. Now combine STEP 2 and STEP 3 to select exactly one of three rigs: MODE STATIC + TYPE D -> RIG 1, ENCLOSED ROUTER. The camera never moves and never leaves the room. MODE STATIC + TYPE A, B or C -> RIG 2, OPEN ROUTER. Interior cameras never move; only the exterior and drone cameras may move. MODE MOVING (any type) -> RIG 3, MOVING ROUTER. Motivated camera movement is unlocked. === THE GOLDEN RULE OF MOVEMENT === Camera movement is never decorative and never random. A camera may move for exactly two reasons: because the subject is moving and the camera is keeping pace with him, or because the shot is an aerial one. In every other case the camera is a locked-off tripod. There is no such thing as a free-floating drift, a random zoom or a sudden reframe in this edit. Every move runs at a constant speed, in a single axis, for the whole duration of its shot, with a gentle ease in and ease out. One shot never contains two different moves. No move ever accelerates, stutters, overshoots or corrects itself. === RIG 1 AND RIG 2 — STATIC SUBJECT (CAM 1 to CAM 12) === CAM 1 — 35mm, eye level, 0 degrees, medium shot waist-up, subject looking directly into this lens. HOME. CAM 2 — 50mm, eye level, 45 degrees right, medium close-up chest-up, three-quarter profile. CAM 3 — 50mm, chest height looking slightly up, 45 degrees left, medium close-up. CAM 4 — 85mm, eye level, 90 degrees right, close-up, clean side profile. CAM 5 — 24mm, raised 70cm above eye level and angled 25 degrees down, 20 degrees right, wide shot showing the full room. CAM 6 — 85mm, slightly above eye level, 70 degrees left, close-up. Reserved for the strongest line. CAM 7 — 24mm, eye level, 25 degrees left, wide master shot at distance. CAM 9 — THE BRIDGE. 28mm, eye level, 60 degrees right, deep wide from the far side of the space with the window and its daylight fully in frame beside the subject, plus a foreground layer in the near field such as a desk edge, monitor, table, cup or chair back. For TYPE C, frame the horizon and the surrounding architecture instead of a window. CAM 10 — 35mm, static, OUTSIDE the location on the other side of the glass, a few metres away, at the same height and floor level, shooting back in. The subject is still recognizable behind the window, the outside world reflects faintly across the glass, the interior reads darker than the exterior. For TYPE C use 85mm from far back with long-lens compression and a soft out-of-focus foreground. CAM 11 — 24mm, drone hovering outside at the subject's level, roughly twelve metres from the facade, performing ONE slow lateral drift at constant speed. The subject is a small figure behind the glass. For TYPE B this may instead be a static position across the street with the full storefront in frame. CAM 12 — 24mm, drone high above and away, performing ONE slow rise at constant speed. The whole building or location and the city around it, the subject no longer identifiable, his voice continuing as voice-over. CAM 1 to CAM 9 are static shots, locked off cameras, tripod-mounted: no dolly, no zoom, no push in, no pull back, no pan, no tilt, no crane, no orbit, no handheld. Inside the location the frame never changes size within a shot. Only CAM 10, CAM 11 and CAM 12 may move. === RIG 3 — MOVING SUBJECT (WCAM 1 to WCAM 9, DRONE 1 to DRONE 3) === Every tracking camera holds a CONSTANT distance from the subject and matches his walking speed exactly. The subject never drifts toward or away from the lens inside a shot unless the shot is WCAM 8. WCAM 1 — 35mm, eye level, directly in front, gimbal travelling backwards ahead of him at exactly his walking speed, constant distance, medium shot waist-up, subject looking into this lens. HOME of the moving mode. WCAM 2 — 50mm, eye level, 45 degrees front-right, tracking alongside at matched speed and constant distance, medium close-up chest-up, three-quarter profile. WCAM 3 — 85mm, eye level, full lateral profile, tracking sideways at matched speed, close-up, foreground elements passing between camera and subject and briefly veiling him. WCAM 4 — 24mm, lowered to knee height, travelling backwards ahead of him at his speed, low wide angle, his feet and the ground surface prominent in the lower frame. WCAM 5 — 50mm, shoulder height, behind and slightly left of the subject, following at matched speed, medium shot over the shoulder revealing what he is walking toward. WCAM 6 — 85mm, eye level, STATIC locked-off camera planted on the ground. The subject walks into frame from a distance, crosses it and passes the lens. The camera itself does not move at all — only the subject moves through the frame. WCAM 7 — 35mm, eye level, ONE slow arc of 40 to 60 degrees around the subject at constant radius and constant speed while he keeps walking, the background sweeping behind him with parallax. WCAM 8 — THE ONLY PUSH-IN IN THE ENTIRE EDIT. 50mm, eye level, travelling ahead of the subject while closing the distance very slightly at constant speed, from medium shot to medium close-up. One single size step and no more. This is a real change of distance, never a zoom, and it happens at most once in the whole video. WCAM 9 — THE BRIDGE. 24mm, eye level, STATIC locked-off wide from across the street or the open space, the subject small in frame walking through the environment, the full location and sky visible around him. This shot must precede the first aerial shot. DRONE 1 — 24mm, drone eight metres above and slightly behind the subject, angled 40 degrees down, following him at his exact walking speed. One constant-speed follow and nothing else. DRONE 2 — 24mm, drone directly overhead looking straight down, performing ONE slow rise at constant speed while he keeps walking below, the street geometry and his shadow opening out beneath him. DRONE 3 — 24mm, drone high above the location, performing ONE slow rise or one slow pull away at constant speed, the subject a tiny figure in the wider city or landscape, his voice continuing as voice-over. STEP 4 — ROUTER. Match the measured duration to one line of the router belonging to your selected rig, and use that camera order exactly, in that order, with no substitutions and no omissions. ENCLOSED ROUTER — RIG 1: 6 to 8 seconds: exactly 4 shots — CAM 1, CAM 2, CAM 5, CAM 7. 9 to 11 seconds: exactly 5 shots — CAM 1, CAM 2, CAM 5, CAM 4, CAM 1. 12 to 14 seconds: exactly 6 shots — CAM 1, CAM 2, CAM 5, CAM 4, CAM 1, CAM 7. 15 to 17 seconds: exactly 7 shots — CAM 1, CAM 2, CAM 5, CAM 4, CAM 1, CAM 6, CAM 7. 18 to 21 seconds: exactly 9 shots — CAM 1, CAM 2, CAM 5, CAM 4, CAM 1, CAM 3, CAM 6, CAM 2, CAM 5. 22 to 25 seconds: exactly 11 shots — CAM 1, CAM 2, CAM 5, CAM 4, CAM 1, CAM 3, CAM 7, CAM 6, CAM 5, CAM 1, CAM 7. 26 to 30 seconds: exactly 13 shots — CAM 1, CAM 2, CAM 5, CAM 4, CAM 1, CAM 3, CAM 7, CAM 2, CAM 5, CAM 6, CAM 4, CAM 3, CAM 7. OPEN ROUTER — RIG 2: 6 to 8 seconds: exactly 4 shots — CAM 1, CAM 2, CAM 9, CAM 12. 9 to 11 seconds: exactly 5 shots — CAM 1, CAM 2, CAM 5, CAM 9, CAM 12. 12 to 14 seconds: exactly 6 shots — CAM 1, CAM 2, CAM 9, CAM 10, CAM 1, CAM 12. 15 to 17 seconds: exactly 7 shots — CAM 1, CAM 2, CAM 9, CAM 10, CAM 6, CAM 1, CAM 12. 18 to 21 seconds: exactly 9 shots — CAM 1, CAM 2, CAM 5, CAM 9, CAM 10, CAM 11, CAM 1, CAM 2, CAM 12. 22 to 25 seconds: exactly 11 shots — CAM 1, CAM 2, CAM 5, CAM 9, CAM 10, CAM 11, CAM 1, CAM 2, CAM 6, CAM 7, CAM 12. 26 to 30 seconds: exactly 13 shots — CAM 1, CAM 2, CAM 5, CAM 4, CAM 1, CAM 9, CAM 10, CAM 11, CAM 1, CAM 6, CAM 2, CAM 5, CAM 12. MOVING ROUTER — RIG 3: 6 to 8 seconds: exactly 4 shots — WCAM 1, WCAM 2, WCAM 9, DRONE 3. 9 to 11 seconds: exactly 5 shots — WCAM 1, WCAM 8, WCAM 6, WCAM 9, DRONE 3. 12 to 14 seconds: exactly 6 shots — WCAM 1, WCAM 8, WCAM 6, WCAM 9, DRONE 1, DRONE 3. 15 to 17 seconds: exactly 7 shots — WCAM 1, WCAM 8, WCAM 6, WCAM 9, DRONE 1, WCAM 1, DRONE 3. 18 to 21 seconds: exactly 9 shots — WCAM 1, WCAM 2, WCAM 6, WCAM 9, DRONE 1, DRONE 2, WCAM 1, WCAM 4, DRONE 3. 22 to 25 seconds: exactly 11 shots — WCAM 1, WCAM 2, WCAM 6, WCAM 5, WCAM 9, DRONE 1, DRONE 2, WCAM 1, WCAM 8, WCAM 4, DRONE 3. 26 to 30 seconds: exactly 13 shots — WCAM 1, WCAM 2, WCAM 6, WCAM 5, WCAM 7, WCAM 9, DRONE 1, DRONE 2, WCAM 1, WCAM 8, WCAM 4, WCAM 5, DRONE 3. STEP 5 — TIMING. Divide the duration across the shots so that the first shot is the longest, wide and aerial shots run slightly longer than close-ups, no shot is shorter than 1.5 seconds and no shot is longer than 4 seconds. Place every cut on a natural speech pause, a sentence end or a breath, never mid-word and never in the middle of a fast head turn or a gesture apex. State each shot's in and out timecode explicitly and close each one with a hard cut. GOING AERIAL: The camera may only leave ground level through the bridge shot — CAM 9 in RIG 2, WCAM 9 in RIG 3 — and never directly from a close-up. Once airborne the shots move outward in a strict ladder: still recognizable, then small in frame, then no longer identifiable. The camera returns to ground level only by cutting back to the home camera, CAM 1 or WCAM 1. The exterior world must match exactly what is visible in @video_1: same location, same time of day, same weather, same quality and direction of daylight, same architecture and same city. Do not invent a different skyline, a different season or a different sky. On aerial shots the subject's face may be small or veiled, so readable lip-sync is not required there, but his body language, walking rhythm and gestures must still match @video_1 at those exact timecodes. MOVEMENT CONTINUITY — RIG 3 ONLY: The subject's walking direction, walking speed and stride are identical to @video_1 at every timecode. He never changes direction, never stops and never starts unless he does so in the source. The environment he walks through is a single continuous space: what is behind him in one shot is what he came from, what is ahead of him is where he is going. Across a cut he must never teleport to a different street, a different building or a different stretch of pavement. Any pedestrians, traffic or background activity behave consistently from shot to shot. EDIT GRAMMAR: Never place the same camera on two consecutive shots. Never place two close-ups back to back. Never follow a moving camera with another moving camera travelling in the opposite direction. Alternate camera-right and camera-left positions to build rhythm. Return to the home camera after every excursion into the air. In RIG 3, at least one shot in every sequence must be a static planted camera — WCAM 6 or WCAM 9 — so that the movement of the others reads as a deliberate choice rather than as constant drift. EYELINE: In @video_1 the subject looks at one fixed point relative to himself — the position the original lens occupied, directly in front of his face. In RIG 1 and RIG 2 that point is fixed in world space and the gaze stays locked on it from every camera, so only CAM 1 reads as direct-to-lens and the off-axis cameras read as three-quarter or profile eyelines. In RIG 3 that point travels with him as he walks, so he reads as direct-to-lens on WCAM 1 and WCAM 8, and as three-quarter, profile or from-behind on the others. He never turns to acknowledge a new camera and never acknowledges the drone. PERFORMANCE: Frame-identical to @video_1 at every timecode. Same gestures, same blinks, same head movements, same micro-expressions, same mouth shapes, same stride. Nothing added, removed, re-timed, looped or frozen. One continuous timeline across the full duration — the cuts change viewpoint only, never the moment in time. The first frame of each shot is the direct temporal continuation of the last frame of the previous shot. OPTICS AND PHYSICS: Perspective, compression and depth of field change correctly with each camera's focal length and distance — 24mm and 28mm show more space and more perspective spread, 85mm compresses and isolates. Motion blur and depth of field respond correctly to each camera's own movement and to the subject's movement. The cameras are physical objects: they never clip through the subject, walls, furniture, glass, vehicles or the ground, and never sit closer than a real lens could. No face warping on the wide lenses. LIGHTING: The light is emitted by the location, not by the camera. Key, fill and rim on the face redistribute correctly for each new angle and shadows fall in the physically correct direction, but the light sources themselves never move, never change colour temperature and never change intensity. Outdoors the sun stays in the same position in the sky for the whole edit. No relighting, no time-of-day shift, no weather change. Exposure and white balance are matched across every camera. AUDIO: Carry the audio track of @video_1 over verbatim, in full, in sync, unchanged. Do not regenerate, do not re-voice, do not re-time, do not denoise, do not add music, do not add sound effects, do not add street noise, wind, footsteps or room tone. Lip movement follows the original recorded audio frame by frame on every ground-level camera. NEGATIVE: no zoom, no digital zoom, no crop-scale, no push in other than the single WCAM 8 shot, no camera movement when the subject is static, no unmotivated drift, no floating camera, no handheld shake, no two consecutive close-ups, no repeating the same camera twice in a row, no cut shorter than 1.5 seconds, no using fewer shots than the router specifies, no cutting from a close-up straight to an aerial shot, no skipping the bridge shot, no camera accelerating or decelerating mid-shot, no two different moves inside one shot, no orbiting drone, no whip pan, no subject changing walking direction or speed, no teleporting to a different street across a cut, no exterior world that contradicts the source, no invented skyline, no camera passing through glass, no new dialogue, no re-voicing, no added music, no added ambience, no lip-sync drift, no face morphing, no identity change, no wardrobe change, no background change, no relighting, no weather change, no time-of-day change, no time jump, no loop, no rewind, no freeze frame, no speed ramp, no slow motion, no dissolve, no transition effect, no text, no captions, no subtitles, no logo, no watermark, no timecode, no letterboxing.

FORMAT: A multi-camera edit of one continuous take. The number of shots and the camera order are NOT chosen freely — they are looked up in the ROUTER below according to the measured duration of @video_1 and the movement mode detected in STEP 2. Camera mov

FORMAT: A multi-camera edit …

Seedance 2.5
ПРОМПТ

FORMAT: A multi-camera edit of one continuous take. The number of shots and the camera order are NOT chosen freely — they are looked up in the ROUTER below according to the measured duration of @video_1 and the movement mode detected in STEP 2. Camera movement is permitted ONLY where this prompt explicitly grants it. Do not use fewer shots than the router specifies. TASK: @video_1 is the source take. Do not re-perform it and do not re-time it. Re-observe the exact same performance from a set of new camera positions, as if a full camera crew had covered the original take simultaneously, and cut between them. The subject's performance, the environment, the wardrobe and the lighting are fixed. Only the point of view changes. STEP 1 — MEASURE THE SOURCE: Read the exact duration of @video_1 in seconds. Read the speech rhythm: sentence boundaries, breaths, pauses, emphasis beats. Read the subject, wardrobe, hair, skin, props, background geometry, lens character, grain and colour science and keep them pixel-consistent in every shot. STEP 2 — DETECT THE MOVEMENT MODE. This decision comes before everything else. Look at whether the subject physically travels through space in @video_1. MODE STATIC — the subject is seated or standing on one spot. His body stays at the same coordinates for the whole take. Only his head, hands and torso move. MODE MOVING — the subject walks, strolls, paces or otherwise travels through space while speaking. His position in the world changes over the duration of the take. If in doubt, choose MODE STATIC. A wrong movement decision is the single most damaging error in this task. STEP 3 — CLASSIFY THE LOCATION. TYPE A — office, business centre, meeting room or lobby with a window wall or large windows. TYPE B — cafe, restaurant, bar or coworking space with a street-facing window. TYPE C — open air: street, pavement, terrace, rooftop, courtyard, park, embankment. TYPE D — enclosed room with no windows and no view outside. Now combine STEP 2 and STEP 3 to select exactly one of three rigs: MODE STATIC + TYPE D -> RIG 1, ENCLOSED ROUTER. The camera never moves and never leaves the room. MODE STATIC + TYPE A, B or C -> RIG 2, OPEN ROUTER. Interior cameras never move; only the exterior and drone cameras may move. MODE MOVING (any type) -> RIG 3, MOVING ROUTER. Motivated camera movement is unlocked. === THE GOLDEN RULE OF MOVEMENT === Camera movement is never decorative and never random. A camera may move for exactly two reasons: because the subject is moving and the camera is keeping pace with him, or because the shot is an aerial one. In every other case the camera is a locked-off tripod. There is no such thing as a free-floating drift, a random zoom or a sudden reframe in this edit. Every move runs at a constant speed, in a single axis, for the whole duration of its shot, with a gentle ease in and ease out. One shot never contains two different moves. No move ever accelerates, stutters, overshoots or corrects itself. === RIG 1 AND RIG 2 — STATIC SUBJECT (CAM 1 to CAM 12) === CAM 1 — 35mm, eye level, 0 degrees, medium shot waist-up, subject looking directly into this lens. HOME. CAM 2 — 50mm, eye level, 45 degrees right, medium close-up chest-up, three-quarter profile. CAM 3 — 50mm, chest height looking slightly up, 45 degrees left, medium close-up. CAM 4 — 85mm, eye level, 90 degrees right, close-up, clean side profile. CAM 5 — 24mm, raised 70cm above eye level and angled 25 degrees down, 20 degrees right, wide shot showing the full room. CAM 6 — 85mm, slightly above eye level, 70 degrees left, close-up. Reserved for the strongest line. CAM 7 — 24mm, eye level, 25 degrees left, wide master shot at distance. CAM 9 — THE BRIDGE. 28mm, eye level, 60 degrees right, deep wide from the far side of the space with the window and its daylight fully in frame beside the subject, plus a foreground layer in the near field such as a desk edge, monitor, table, cup or chair back. For TYPE C, frame the horizon and the surrounding architecture instead of a window. CAM 10 — 35mm, static, OUTSIDE the location on the other side of the glass, a few metres away, at the same height and floor level, shooting back in. The subject is still recognizable behind the window, the outside world reflects faintly across the glass, the interior reads darker than the exterior. For TYPE C use 85mm from far back with long-lens compression and a soft out-of-focus foreground. CAM 11 — 24mm, drone hovering outside at the subject's level, roughly twelve metres from the facade, performing ONE slow lateral drift at constant speed. The subject is a small figure behind the glass. For TYPE B this may instead be a static position across the street with the full storefront in frame. CAM 12 — 24mm, drone high above and away, performing ONE slow rise at constant speed. The whole building or location and the city around it, the subject no longer identifiable, his voice continuing as voice-over. CAM 1 to CAM 9 are static shots, locked off cameras, tripod-mounted: no dolly, no zoom, no push in, no pull back, no pan, no tilt, no crane, no orbit, no handheld. Inside the location the frame never changes size within a shot. Only CAM 10, CAM 11 and CAM 12 may move. === RIG 3 — MOVING SUBJECT (WCAM 1 to WCAM 9, DRONE 1 to DRONE 3) === Every tracking camera holds a CONSTANT distance from the subject and matches his walking speed exactly. The subject never drifts toward or away from the lens inside a shot unless the shot is WCAM 8. WCAM 1 — 35mm, eye level, directly in front, gimbal travelling backwards ahead of him at exactly his walking speed, constant distance, medium shot waist-up, subject looking into this lens. HOME of the moving mode. WCAM 2 — 50mm, eye level, 45 degrees front-right, tracking alongside at matched speed and constant distance, medium close-up chest-up, three-quarter profile. WCAM 3 — 85mm, eye level, full lateral profile, tracking sideways at matched speed, close-up, foreground elements passing between camera and subject and briefly veiling him. WCAM 4 — 24mm, lowered to knee height, travelling backwards ahead of him at his speed, low wide angle, his feet and the ground surface prominent in the lower frame. WCAM 5 — 50mm, shoulder height, behind and slightly left of the subject, following at matched speed, medium shot over the shoulder revealing what he is walking toward. WCAM 6 — 85mm, eye level, STATIC locked-off camera planted on the ground. The subject walks into frame from a distance, crosses it and passes the lens. The camera itself does not move at all — only the subject moves through the frame. WCAM 7 — 35mm, eye level, ONE slow arc of 40 to 60 degrees around the subject at constant radius and constant speed while he keeps walking, the background sweeping behind him with parallax. WCAM 8 — THE ONLY PUSH-IN IN THE ENTIRE EDIT. 50mm, eye level, travelling ahead of the subject while closing the distance very slightly at constant speed, from medium shot to medium close-up. One single size step and no more. This is a real change of distance, never a zoom, and it happens at most once in the whole video. WCAM 9 — THE BRIDGE. 24mm, eye level, STATIC locked-off wide from across the street or the open space, the subject small in frame walking through the environment, the full location and sky visible around him. This shot must precede the first aerial shot. DRONE 1 — 24mm, drone eight metres above and slightly behind the subject, angled 40 degrees down, following him at his exact walking speed. One constant-speed follow and nothing else. DRONE 2 — 24mm, drone directly overhead looking straight down, performing ONE slow rise at constant speed while he keeps walking below, the street geometry and his shadow opening out beneath him. DRONE 3 — 24mm, drone high above the location, performing ONE slow rise or one slow pull away at constant speed, the subject a tiny figure in the wider city or landscape, his voice continuing as voice-over. STEP 4 — ROUTER. Match the measured duration to one line of the router belonging to your selected rig, and use that camera order exactly, in that order, with no substitutions and no omissions. ENCLOSED ROUTER — RIG 1: 6 to 8 seconds: exactly 4 shots — CAM 1, CAM 2, CAM 5, CAM 7. 9 to 11 seconds: exactly 5 shots — CAM 1, CAM 2, CAM 5, CAM 4, CAM 1. 12 to 14 seconds: exactly 6 shots — CAM 1, CAM 2, CAM 5, CAM 4, CAM 1, CAM 7. 15 to 17 seconds: exactly 7 shots — CAM 1, CAM 2, CAM 5, CAM 4, CAM 1, CAM 6, CAM 7. 18 to 21 seconds: exactly 9 shots — CAM 1, CAM 2, CAM 5, CAM 4, CAM 1, CAM 3, CAM 6, CAM 2, CAM 5. 22 to 25 seconds: exactly 11 shots — CAM 1, CAM 2, CAM 5, CAM 4, CAM 1, CAM 3, CAM 7, CAM 6, CAM 5, CAM 1, CAM 7. 26 to 30 seconds: exactly 13 shots — CAM 1, CAM 2, CAM 5, CAM 4, CAM 1, CAM 3, CAM 7, CAM 2, CAM 5, CAM 6, CAM 4, CAM 3, CAM 7. OPEN ROUTER — RIG 2: 6 to 8 seconds: exactly 4 shots — CAM 1, CAM 2, CAM 9, CAM 12. 9 to 11 seconds: exactly 5 shots — CAM 1, CAM 2, CAM 5, CAM 9, CAM 12. 12 to 14 seconds: exactly 6 shots — CAM 1, CAM 2, CAM 9, CAM 10, CAM 1, CAM 12. 15 to 17 seconds: exactly 7 shots — CAM 1, CAM 2, CAM 9, CAM 10, CAM 6, CAM 1, CAM 12. 18 to 21 seconds: exactly 9 shots — CAM 1, CAM 2, CAM 5, CAM 9, CAM 10, CAM 11, CAM 1, CAM 2, CAM 12. 22 to 25 seconds: exactly 11 shots — CAM 1, CAM 2, CAM 5, CAM 9, CAM 10, CAM 11, CAM 1, CAM 2, CAM 6, CAM 7, CAM 12. 26 to 30 seconds: exactly 13 shots — CAM 1, CAM 2, CAM 5, CAM 4, CAM 1, CAM 9, CAM 10, CAM 11, CAM 1, CAM 6, CAM 2, CAM 5, CAM 12. MOVING ROUTER — RIG 3: 6 to 8 seconds: exactly 4 shots — WCAM 1, WCAM 2, WCAM 9, DRONE 3. 9 to 11 seconds: exactly 5 shots — WCAM 1, WCAM 8, WCAM 6, WCAM 9, DRONE 3. 12 to 14 seconds: exactly 6 shots — WCAM 1, WCAM 8, WCAM 6, WCAM 9, DRONE 1, DRONE 3. 15 to 17 seconds: exactly 7 shots — WCAM 1, WCAM 8, WCAM 6, WCAM 9, DRONE 1, WCAM 1, DRONE 3. 18 to 21 seconds: exactly 9 shots — WCAM 1, WCAM 2, WCAM 6, WCAM 9, DRONE 1, DRONE 2, WCAM 1, WCAM 4, DRONE 3. 22 to 25 seconds: exactly 11 shots — WCAM 1, WCAM 2, WCAM 6, WCAM 5, WCAM 9, DRONE 1, DRONE 2, WCAM 1, WCAM 8, WCAM 4, DRONE 3. 26 to 30 seconds: exactly 13 shots — WCAM 1, WCAM 2, WCAM 6, WCAM 5, WCAM 7, WCAM 9, DRONE 1, DRONE 2, WCAM 1, WCAM 8, WCAM 4, WCAM 5, DRONE 3. STEP 5 — TIMING. Divide the duration across the shots so that the first shot is the longest, wide and aerial shots run slightly longer than close-ups, no shot is shorter than 1.5 seconds and no shot is longer than 4 seconds. Place every cut on a natural speech pause, a sentence end or a breath, never mid-word and never in the middle of a fast head turn or a gesture apex. State each shot's in and out timecode explicitly and close each one with a hard cut. GOING AERIAL: The camera may only leave ground level through the bridge shot — CAM 9 in RIG 2, WCAM 9 in RIG 3 — and never directly from a close-up. Once airborne the shots move outward in a strict ladder: still recognizable, then small in frame, then no longer identifiable. The camera returns to ground level only by cutting back to the home camera, CAM 1 or WCAM 1. The exterior world must match exactly what is visible in @video_1: same location, same time of day, same weather, same quality and direction of daylight, same architecture and same city. Do not invent a different skyline, a different season or a different sky. On aerial shots the subject's face may be small or veiled, so readable lip-sync is not required there, but his body language, walking rhythm and gestures must still match @video_1 at those exact timecodes. MOVEMENT CONTINUITY — RIG 3 ONLY: The subject's walking direction, walking speed and stride are identical to @video_1 at every timecode. He never changes direction, never stops and never starts unless he does so in the source. The environment he walks through is a single continuous space: what is behind him in one shot is what he came from, what is ahead of him is where he is going. Across a cut he must never teleport to a different street, a different building or a different stretch of pavement. Any pedestrians, traffic or background activity behave consistently from shot to shot. EDIT GRAMMAR: Never place the same camera on two consecutive shots. Never place two close-ups back to back. Never follow a moving camera with another moving camera travelling in the opposite direction. Alternate camera-right and camera-left positions to build rhythm. Return to the home camera after every excursion into the air. In RIG 3, at least one shot in every sequence must be a static planted camera — WCAM 6 or WCAM 9 — so that the movement of the others reads as a deliberate choice rather than as constant drift. EYELINE: In @video_1 the subject looks at one fixed point relative to himself — the position the original lens occupied, directly in front of his face. In RIG 1 and RIG 2 that point is fixed in world space and the gaze stays locked on it from every camera, so only CAM 1 reads as direct-to-lens and the off-axis cameras read as three-quarter or profile eyelines. In RIG 3 that point travels with him as he walks, so he reads as direct-to-lens on WCAM 1 and WCAM 8, and as three-quarter, profile or from-behind on the others. He never turns to acknowledge a new camera and never acknowledges the drone. PERFORMANCE: Frame-identical to @video_1 at every timecode. Same gestures, same blinks, same head movements, same micro-expressions, same mouth shapes, same stride. Nothing added, removed, re-timed, looped or frozen. One continuous timeline across the full duration — the cuts change viewpoint only, never the moment in time. The first frame of each shot is the direct temporal continuation of the last frame of the previous shot. OPTICS AND PHYSICS: Perspective, compression and depth of field change correctly with each camera's focal length and distance — 24mm and 28mm show more space and more perspective spread, 85mm compresses and isolates. Motion blur and depth of field respond correctly to each camera's own movement and to the subject's movement. The cameras are physical objects: they never clip through the subject, walls, furniture, glass, vehicles or the ground, and never sit closer than a real lens could. No face warping on the wide lenses. LIGHTING: The light is emitted by the location, not by the camera. Key, fill and rim on the face redistribute correctly for each new angle and shadows fall in the physically correct direction, but the light sources themselves never move, never change colour temperature and never change intensity. Outdoors the sun stays in the same position in the sky for the whole edit. No relighting, no time-of-day shift, no weather change. Exposure and white balance are matched across every camera. AUDIO: Carry the audio track of @video_1 over verbatim, in full, in sync, unchanged. Do not regenerate, do not re-voice, do not re-time, do not denoise, do not add music, do not add sound effects, do not add street noise, wind, footsteps or room tone. Lip movement follows the original recorded audio frame by frame on every ground-level camera. NEGATIVE: no zoom, no digital zoom, no crop-scale, no push in other than the single WCAM 8 shot, no camera movement when the subject is static, no unmotivated drift, no floating camera, no handheld shake, no two consecutive close-ups, no repeating the same camera twice in a row, no cut shorter than 1.5 seconds, no using fewer shots than the router specifies, no cutting from a close-up straight to an aerial shot, no skipping the bridge shot, no camera accelerating or decelerating mid-shot, no two different moves inside one shot, no orbiting drone, no whip pan, no subject changing walking direction or speed, no teleporting to a different street across a cut, no exterior world that contradicts the source, no invented skyline, no camera passing through glass, no new dialogue, no re-voicing, no added music, no added ambience, no lip-sync drift, no face morphing, no identity change, no wardrobe change, no background change, no relighting, no weather change, no time-of-day change, no time jump, no loop, no rewind, no freeze frame, no speed ramp, no slow motion, no dissolve, no transition effect, no text, no captions, no subtitles, no logo, no watermark, no timecode, no letterboxing.

UNISET

ИСТОРИЯ