Establishing the walk, shot 1
From MiniMax H3 long takes from chained renders

Watch it in the video at 2:53
- Resolution
- 1376×768
- Length
- 243 frames, 10.12s
- Handoff
- First render, nothing handed in
- Steps
- 20
- Seed
- 1078725334536121
- Ran on
- RTX 5090 on RunPod
Rent the same kind of GPU on RunPodAffiliate link: Loop Forge gets credit if you sign up through it, at no extra cost to you.
subject_definitions: <Subject 1> is the young woman in <Picture 2>, in her early twenties, with long thick copper-orange curly hair falling well past her shoulders, pale freckled skin, blue-grey eyes and a slim build, wearing a plain cream short-sleeved crew-neck t-shirt, a long dark iron cross pendant on a thin dark chain, blue denim jeans, a wide tan leather belt with a brass buckle, a small tan leather belt pouch on her right hip, dark braided rope bracelets on both wrists and thin silver rings. Only her face, hair, body and clothing are taken from <Picture 2>; the snow, cliffs, green aurora and wooden sailing ship behind her in that image are not part of the target video. <Subject 2> is the block of Market Street in San Francisco photographed from directly overhead in <Picture 1>: a broad straight avenue cutting diagonally between dense financial-district towers, sheer dark glass curtain-wall high-rises along the left side, a slim pale stone tower and a brick-and-terracotta building with an octagonal corner turret and a green copper cupola on the right, wide brick-red paved sidewalks, white-striped crosswalks at the intersection and a small sunken octagonal planter set into the brick paving. Only the street layout, building shapes, materials and colours are used; the overhead aerial viewpoint and map-like flattening of <Picture 1> are not retained, and the target video is filmed at street level. summary: [reference generation] The target video opens the continuous front-tracking walk: <Subject 1> walks north-east along the Market Street block of <Subject 2> in a single unbroken fifteen-second take, coming off a crosswalk, settling into her stride, checking her phone and putting it away as the camera retreats ahead of her at her own pace. retention_analysis: <Subject 1> (appears in [Shot 1]): fully_preserved - her face, freckles, blue-grey eyes, long copper-orange curls, cream t-shirt, iron cross pendant, tan leather belt and belt pouch, denim jeans, braided rope bracelets and silver rings are retained exactly. <Subject 2> (appears in [Shot 1]): partially_preserved - the dark glass towers, the slim pale stone tower, the turreted brick building with its green copper cupola, the brick-red paving, the crosswalk stripes and the sunken octagonal planter are retained; the overhead aerial viewpoint is replaced by a street-level camera at chest height. detailed_description: The target video is photorealistic live-action in a naturalistic documentary-cinematic style, filmed on a stabilised gimbal with a 35mm-equivalent lens held at chest height, in clear late-afternoon San Francisco light, with no on-screen text and no subtitles anywhere. [Shot 1] The camera sits in front of <Subject 1> and faces back toward her in one continuous fifteen-second take, matching her speed exactly so her size and position in the frame stay constant; she is centred and framed from mid-thigh up, and there is no cut at any point. The camera holds its distance and never pushes in, pulls out or zooms. When she slows or stops, the camera slows and stops with her, holding the same gap, and picks up again at the moment she does. <Subject 1>, the young woman with long thick copper-orange curls, pale freckled skin, a cream crew-neck t-shirt, the long dark iron cross pendant, a wide tan leather belt with a small tan pouch on her right hip and blue denim jeans, walks directly toward the lens along the brick-red paving of <Subject 2>. She has just stepped off a white-striped crosswalk, and over the first three seconds her stride settles from the quicker crossing pace into an even, unhurried walk, arms swinging loosely, rope bracelets shifting on her wrists. Between the scripted beats her face is relaxed and unselfconscious, the ordinary expression of someone walking a route she knows well: lips closed, jaw loose, eyes moving over the buildings, the traffic and the pavement ahead, narrowing now and then against the low sun. She does not smile at the lens and does not perform for the camera. Any change of expression is a brief, specific reaction to something in the scene and settles back to that neutral walking face within a second or two. The low late-afternoon sun sits behind her and slightly to her left, far up the avenue, rimming her curls and shoulders in warm orange and throwing her long shadow forward past the camera; the lens faces into that light with a soft gentle flare across the upper frame while her face stays readable in cool bounced light from the pale stone tower opposite. Behind her the avenue recedes in a deep perspective of dark glass curtain-wall towers, and the brick-and-terracotta building with its octagonal turret and green copper cupola stands to her right, its upper floors catching the last direct sun. Around the five-second mark a gust off the street blows her hair across her face; she lifts her right hand and sweeps it back behind her ear, tilting her chin up briefly to take in the tops of the buildings before her gaze drops level again. A city bus pulls away behind her and she tracks it for a moment without turning her head. Near the nine-second mark she reaches into the small tan belt pouch, draws out a plain dark phone with no visible brand mark, glances at the screen for two seconds with her brow faintly drawn, then slides it back into the pouch and presses the flap flat without breaking stride. Other pedestrians in coats and work clothes cross the frame in soft focus at the edges, walking the other way, and the sunken octagonal planter passes at the right edge. Through the final two seconds she is centred, mid-stride, walking at the same steady pace with that neutral, unhurried expression. overall_soundscape: Steady open-air city ambience fills the whole segment: distant traffic on the avenue, a bus pulling away, the layered murmur of passing pedestrians and a faint wind moving between the towers. Her footsteps fall in an even, unhurried rhythm on the brick paving throughout, with the light rustle of denim and the small click of the belt-pouch flap close to the microphone. non_diegetic_music: N/A
Notes
She comes off a crosswalk and settles into stride

