STEP 1, the prompt Claude wrote for the talking take (Template A)

One single continuous 5-second 16:9 landscape video. #video_1 is the source: one continuous take of a person talking to camera, filmed from one fixed camera only. Re-shoot this exact take as a multi-camera edit.

VOICE AND TIMING. #video_1 is the master for the timing, the audio and every word. Keep the audio of #video_1 exactly: the same voice, the same words, the same pace, spoken once, never restarting. Do not summarise, shorten, rewrite, replace, mute or invent any dialogue. Their lips match the audio of #video_1 at every moment, in every angle. Real-time playback: no speed change, no retiming.

PERSON AND ROOM. Keep everything from #video_1 exactly: the same person, face, hair, skin, clothing and body, a light heather-grey pullover hoodie with a small black print on the chest, the same room with a plain cream wall, a tall window with a pale curtain and a stack of colourful books on a small wooden shelf to the left, a dark grey gaming chair behind him on the right, and a black broadcast microphone on a stand in the lower left of the frame, the same light. The lights do not move with the camera: shadows stay where the room's lights put them in every shot. The ONLY thing that changes between cuts is where the camera is.

CUTS. Instant hard cuts only, and only in the natural pauses and breaths between phrases, never in the middle of a word. Spread the cuts evenly across the take.
CUT 1: the original camera, exactly as #video_1.
CUT 2 (from 2.1 seconds, the breath just before his first "this side"): a true 90-degree side profile at eye level from the window side of the room (the left of the original frame), medium close-up, the camera exactly perpendicular to the way he faces, looking across him to the plain cream wall; he keeps talking towards the original camera.
CUT 3 (from 3.9 seconds, the breath just before his second "this side"): a high three-quarter angle roughly 40 degrees above eye level from the plain-wall side of the room (the right of the original frame), about 35 degrees off centre, tilted down, medium shot; he keeps talking towards the original camera.
Every shot is visibly different from the one before it. No two neighbouring shots look the same. No shot is a zoomed-in or cropped version of another: every cut puts the camera in a new position in the room. In every shot the full head stays in frame with room above the hair.

NOT ALLOWED. No transitions, no dissolves, no zooms, no camera movement within a shot, no new background, furniture or props, no visible cameras or rigs, no text, captions or logos.

Make this ONE single Seedance 2.5 generation of the whole prompt exactly as written: no separate scenes, no assembly step.


STEP 3, the sentence that asks for a draft first (end of the prompt)

Generate it in Seedance 2.5 draft mode first: a 480p draft preview, so I can approve it and then finish that same draft at full resolution from its draft ID. Do not generate the full resolution version yet.


STEP 3, the line that finishes the draft, same chat

That draft is the one I want. Finish that exact draft at full quality from its draft ID: the same take, the same shots, the same timing and the same sound, rendered as the final full quality Seedance 2.5 video. Do not generate a new video and do not change anything in it.


STEP 4, CCTV to cinema (Template B, draft first)

One single continuous 10-second 16:9 landscape video. #video_1 is the source: one continuous shot of a purple double-decker bus mounting the pavement outside a pub, hitting a man and throwing him to the ground before he gets back up, filmed from one high security camera only. Re-shoot this exact moment as a multi-camera edit.

OBSERVE, DO NOT ACT. Nothing new is acted or performed: this is the same moment that already happened in #video_1, observed from new camera positions. The bus, the man, the pub front, the lamp post and the debris are fixed in space exactly where #video_1 has them; only the camera moves.

TIMING AND ACTION. #video_1 is the master for the timing, the action and the sound. Every movement happens exactly as in #video_1, at the same moment and the same speed: from 0 seconds the bus comes down the road towards the pub; at 3.0 seconds it mounts the pavement and smashes through the pub's wooden frontage, timber flying; at 3.5 seconds it hits the man on the pavement and throws him to the ground; from 4.5 seconds the bus sweeps away to the left and a cloud of dust fills the street; from 6 seconds the man is on his hands and knees on the pavement in the dust; from 8.5 seconds he gets up and stands by the purple pub front. The timing of every movement, the impact and the fall is locked to #video_1. Keep the sound of #video_1. Real-time playback: no slow motion, no speed change, no retiming, no extra action, nothing happens that does not happen in #video_1.

PEOPLE AND PLACE. Keep everything from #video_1 exactly: the bus is a purple double-decker with "Reading" on its side; the man is bald, in a black waistcoat over a white shirt and light blue jeans; the place is a street corner in an English town, a two-lane road with double yellow lines, a pub with a purple painted front on the corner, a lamp post, brick buildings up the road, overcast daylight. The light does not move with the camera. The new camera angles are clean, sharp cinematic footage: no timestamp, no fisheye, no low resolution, no CCTV look. The ONLY thing that changes between cuts is where the camera is.

CUTS. 3 shots only, instant hard cuts on action beats, so every new angle catches the same action already in progress. Every camera stays near the original camera's line and faces the same way, up the road.
SHOT 1 (0 to 3.2 seconds): lowered to street level on the pavement outside the pub, the lens at head height, facing up the road as the bus comes towards the camera.
SHOT 2 (3.2 to 6.0 seconds): shifted to the left of the original camera position, eye level, medium shot, as the bus smashes the frontage and the man is thrown to the ground, timber and dust flying.
SHOT 3 (6.0 to 10.0 seconds): closer along the original camera's line, a medium close-up of the man on his hands and knees in the dust as he gets back up.
The original camera angle of #video_1 never appears, in any shot, including the final second.
Every shot is visibly different from the one before it. No two neighbouring shots look the same.

NOT ALLOWED. No transitions, no dissolves, no zooms, no camera movement within a shot, no new people, vehicles, objects or props, no visible cameras or rigs, no added text, captions, timestamps or logos.

Make this ONE single Seedance 2.5 generation of the whole prompt exactly as written: no separate scenes, no assembly step. Generate it in Seedance 2.5 draft mode first: a 480p draft preview, so I can approve it and then finish that same draft at full resolution from its draft ID. Do not generate the full resolution version yet.


STEP 4, a product shot (Template B)

One single continuous 7-second 16:9 landscape video. #video_1 is the source: one continuous locked-off product shot of a hand pouring cold brew coffee from a small glass jug into a tall glass of milk and ice, filmed from one camera only. Re-shoot this exact moment as a multi-camera product ad.

OBSERVE, DO NOT ACT. Nothing new is acted or performed: this is the same moment that already happened in #video_1, observed from new camera positions. The glass, the jug, the hands, the milk bottle and the board are fixed in space exactly where #video_1 has them; only the camera moves.

TIMING AND ACTION. #video_1 is the master for the timing and the action. Every movement happens exactly as in #video_1, at the same moment and the same speed: at 1 second a hand brings the jug of dark cold brew in from above; from 1.5 to 3.5 seconds it pours the coffee over the ice into the milk; at 3.5 seconds the jug lifts away out of frame; from 4 seconds the coffee clouds down through the milk while the other hand rests on the table. The timing of the pour is locked to #video_1. Real-time playback: no slow motion, no speed change, no retiming, no extra action, nothing happens that does not happen in #video_1.

PEOPLE AND PLACE. Keep everything from #video_1 exactly: a tall clear glass full of ice and milk that ALREADY holds a first swirl of coffee from an earlier pour, as #video_1 shows from its first frame (light coffee-brown clouds around the ice in the top third, white milk below); a small clear glass jug of dark cold brew coffee; a swing-top glass bottle of milk standing to the right of the glass; a grey marble board with a metal straw on the left; a pale pink and white background; soft daylight from the left. The light does not move with the camera. The ONLY thing that changes between cuts is where the camera is.

CUTS. 3 shots only, instant hard cuts on action beats, so every new angle catches the same action already in progress. Every camera stays on the original camera's side of the product and faces the same way.
SHOT 1 (0 to 2.5 seconds): lowered below table height, the lens resting on the table surface, looking up at the glass so the glass towers over the camera and the wall is visible above it, a low hero angle, the jug tilting in from above and the new stream of dark coffee hitting the ice at 1.5 seconds, the earlier swirl of coffee already visible around the ice exactly as in #video_1.
SHOT 2 (2.5 to 4.5 seconds): a high angle about 40 degrees above the table, tilted down, looking down into the top of the glass as the coffee pours over the ice and the jug lifts away at 3.5 seconds.
SHOT 3 (4.5 to 7 seconds): shifted well to the right of the original camera position, about 45 degrees around the table, at the height of the glass, a three-quarter angle with the milk bottle standing in front of the glass between the camera and the glass, large and soft in the foreground, the glass sharp behind it, as the coffee clouds through the milk.
The original camera angle of #video_1 never appears, in any shot, including the final second. None of the three shots matches the framing of #video_1.
Every shot is visibly different from the one before it. No two neighbouring shots look the same. No shot is a zoomed-in or cropped version of another: every cut puts the camera in a new position. The glass stays fully in frame in every shot.

NOT ALLOWED. No transitions, no dissolves, no zooms, no camera movement within a shot, no new people, objects or props, no visible cameras or rigs, no text, captions or logos.

Make this ONE single Seedance 2.5 generation of the whole prompt exactly as written: no separate scenes, no assembly step.