Mark H

Unlearning, one cut at a time.

[Mark H @markh_app]

Seeddance 2.0 mini prompt:
Super casual real smartphone home video footage, a road trip vlog filmed by an off-camera friend, natural mobile phone camera with slight authentic handheld shake, normal frame rate with smooth natural motion, rapidfire montage with quick jump cuts every 1-2 seconds, unpolished authentic phone recording, pure raw home video feel, no cinematic polish, 16:9 landscape.

[IDENTITY LOCK] The provided first frame is the strict and ONLY reference for the woman. Lock her facial features (eye shape, eyebrow shape, nose shape, lip shape, jaw and bone structure, skin tone), her hairstyle, and everything she is wearing, with zero deviation in every shot. Her clothing and accessories must remain exactly identical across all cuts - same garments, same colours, nothing added, nothing removed, no change of outfit at any point. Do not add any garment, jacket, scarf or accessory that is not visible in the first frame. Body proportions and silhouette are capped by the first frame, not by text. Head angle, face orientation, gaze and expression are free and controlled by the shot descriptions below. Her face must stay identical across every cut.

[BOTTOM] Her lower body is not visible in the first frame. In any shot where it becomes visible she wears plain black short fitted athletic shorts, unchanged in every shot.

[CAST] Exactly ONE person visible in frame at all times. No bystanders, no other people, no reflections of other people. The camera is held by an unseen friend who never appears.

[VEHICLE] A large angular pickup truck with flat brushed stainless-steel panels and sharp straight edges. Never show the whole vehicle. It appears only as partial panels, a door, a fender, a wheel, an open door frame, or a reflective surface at the edge of frame.

[LOCATION] Every exterior shot is in mainland China. Chinese urban and roadside environments only, with Chinese street signage, Chinese shopfront lettering and Chinese highway markings. Never American or European streets.

0-2.5s: Continue directly from the first frame. She stands beside the truck on the city street, smiling and turning slightly, wind moving her hair, handheld camera drifting.
2.5-5s: Abrupt cut to a wide plate of a dense Chinese city skyline and elevated expressway, Chinese signage on the buildings, no vehicle in frame, camera drifting slightly.
5-7.5s: Jump cut to a charging station, a row of tall white charging posts with red accents and Chinese station signage. Charging is finished. She walks the last step toward the nearest post holding a thick black cable loosely down at her side, hand near her hip, and reaches up to hook it back onto the post. Medium shot from a few steps away, her upper body and the post both in frame, the cable soft and out of focus, her face turning toward the camera with a relaxed smile as she finishes.
7.5-10s: Fast shaky handheld from outside the vehicle, the driver door is open and she is sitting sideways in the driver seat eating a snack, feet resting on the door sill, the angular stainless steel door and interior visible around her, charging posts behind the camera side.
10-12.5s: Quick cut to a moving shot through the windshield from the passenger side, green hills and open Chinese highway ahead, she glances over with a playful smile, hair moving in the airflow.
12.5-15s: Final rapid transition at dusk, close on her leaning back against a closed door panel, soft calm smile, warm evening light, gentle phone sway.

Natural smartphone video quality, slight handheld shake, smooth motion, authentic physics, matte natural skin, no beauty filter, stable single-character consistency, identical outfit throughout, no pro effects, no on-screen text, no logos on the vehicle body.

[Mark H @markh_app]

真的躺着就横跨中国了 全程 FSD 车自己开 😏
Road trip across China. The car drove, I ate.

Prompt 👇

[Mark H @markh_app]

Seeddance 2.0 prompt:

Style: 8K cinematic photorealistic image. Real photographic quality. No 3D rendering, no game engine feel, no game cutscene aesthetic.

Cinematography: Naturalistic cinematography, like a real in-car selfie video polished to a high-end quality.

Lighting: Only daytime natural diffused light, soft, even, low contrast, with main light coming from the sky and the car windows. Light on the subject's face is flat and soft, with balanced illumination on both cheeks, no hard-edged shadows, no strong contrast between one bright side and one dark side. No studio light, no studio lighting setup, no fake AI light, no waxy highlights.

Color: 60:30:10 color scheme. Primary is the black car interior (black leather seats, black roof and trim). Secondary is natural skin tone, white spaghetti-strap top, and off-white plush toy body. Accent is the warm champagne-gold door frames, black seatbelt, and light gray-brown of the plush toy's inner ears and paw pads.

Lens: Real physical lens, slight wide-angle, 180-degree shutter motion blur. The camera position is completely locked for the entire piece. Absolutely no push-in/pull-out, no zoom (no zoom in/out), no left/right pan, no up/down tilt. Every shot's internal composition stays completely static from start to finish. Only people and objects within the frame move, the camera itself does not.

Skin: Real skin texture with pores, fine peach fuzz, natural blood-tone coloring and real light interaction. No plastic skin.

Performance: High-end natural performance. Eyes have moments of stillness, breathing and chest movement are natural, eyes have real reflections.

Physics: All objects have real weight. The plush toy has real fluffy softness and realistic compression/deformation. The seatbelt hugs the body. Food and steam have realistic physical behavior. Interior shadows are realistic.

Composition: Fixed in-car camera position, rule-of-thirds composition. The subject is in motion from the first frame, but the camera itself never moves.

Continuity: In every shot, the subject's face, hairstyle, clothing, plush toy design and color, and interior environment must remain 100% consistent with the reference image (first frame). Do not regenerate or alter the subject's face, do not face-swap, do not change the plush toy's design or color.

Technical: 24fps smooth motion, 8K detail, no jitter or glitches.

Audio: Ambient sound only, no music, no subtitles.

Subject — 100% match to the person and plush toy in the reference first-frame image. Do not use any selfie reference tag or additional reference images. All identity information comes from the first-frame image. Adult female sitting in the right-side passenger seat of the vehicle. Facial features, face shape, hairstyle, hair color, and overall vibe strictly follow the first frame. Do not alter her appearance. Beside her, on the left seat, is the same giant original plush toy — an original design between a rabbit and an elf: off-white long plush fur, head slightly larger than body, two long upright pointed ears, a pair of large round dark-brown button eyes, mouth showing two small pointed fangs, with a cute yet slightly mischievous expression. Its body nearly fills the entire left seat, with its head near the ceiling. The seatbelt is fastened across it. The plush toy always stays within its own seat area and never crosses over. White balance 6500K.

Vehicle — Tesla Cybercab, a two-seat fully autonomous driverless vehicle. There is no steering wheel, no driver's seat, and no driving controls of any kind in the cabin. Black leather seats, black roof and door frames, warm champagne-gold body visible around the door frames. All interior panels, armrests, and console are black. No silver or shiny surfaces.

Location — The vehicle is driving or briefly stopped on a daytime Shanghai city street. Mature plane trees on both sides, mid-rise residential buildings, street-level shops, non-motor-vehicle lane. Because Shanghai roads are relatively narrow, objects outside the window are close to the glass, not the wide open American-style distant landscape. No Oriental Pearl Tower, no Lujiazui skyline, no futuristic city.

Action — Fixed-camera hard-cut montage, 4 hard-cut shots showing the "only on Cybercab" theme: because it's fully autonomous, no one needs to drive, so inside the car you can do anything you couldn't normally do while driving. Throughout, no one touches any driving controls and the vehicle is always in autonomous mode. All shots have a completely fixed camera position — no push, no pull, no zoom, no pan.

Shot 1 (0:00–0:03) — Continuation of the first-frame image: She and the giant plush toy are already sitting side by side in the moving Cybercab, seatbelts fastened on both. The cabin has no steering wheel and no driver's seat, emphasizing the empty "fully driverless" front cabin. The street scene outside moves smoothly backward, showing the vehicle in normal motion. Her expression is relaxed and natural, enjoying the pure-passenger experience. Hard cut.

Shot 2 (0:03–0:06) — Hard cut to her making a subtle, elegant, coquettish small movement to music: one hand lifts gently to her cheek or beside her head, fingers naturally curved with a soft delicate motion at the fingertips, eyes softly closed, corners of her mouth showing a faint, contented smile. Her head and upper body only sway and tilt in an extremely subtle, graceful way — not any large motion or exaggerated dance. Overall vibe is languid, sensual yet restrained, as if she's casually humming along in her seat with tiny movements, not dancing. Seatbelt stays fastened. Plush toy stays in its original seated position. Hard cut.

Shot 3 (0:06–0:09) — Hard cut to her holding a sheng jian bao (pan-fried bun) or xiao long bao (soup dumpling) in her hand, taking a small bite. The motion is simple and natural. Occasionally the steam wafts across her face slightly. Her expression is content and easygoing. Her other hand naturally rests on her knee or on the packaging. No complex props or large movements needed. The plush toy sits quietly beside her as if "keeping her company." The camera position stays absolutely fixed throughout — no movement of any kind. Hard cut.

Shot 4 (0:09–0:12) — Hard cut to a quiet frame: her whole body leans sideways toward the giant plush toy on her left, head resting gently against the plush toy's soft body or shoulder. Her eyes close naturally. Her face carries no smile and no expression at all, showing a real, peaceful sleeping state. Breathing and chest movement are natural. The seatbelt remains fastened. The whole frame is quiet and natural, as if she really has fallen asleep — not a posed, sweet-looking expression. Hard cut.

Camera notes — Shot 1: Fixed in-car wide-angle camera position, capturing the full view of subject and plush toy sitting side by side in the driverless cabin. Shot 2: Same fixed camera, capturing the delicate, elegant small motion. Camera never moves or zooms. Shot 3: Same fixed camera, capturing the simple eating motion with the sheng jian bao / xiao long bao. Camera never moves, no push/pull, no zoom. Shot 4: Same fixed camera. The overall tone shifts to quiet, focusing on the moment of her falling asleep leaning against the plush toy. No cinematic push/pull, no drone shots, no floating camera, no zoom of any kind for the entire piece.

Avoid: handheld selfie feel, steering wheel, driver's seat, driving posture, someone driving, face swap, plush toy design or color changes, plush toy crossing over into the female subject's seat, plush toy shrinking in size, silver trim, silver console, gray seats, fabric seats, high-contrast harsh light on the face, studio light, studio lighting, hard light, fake AI light, waxy highlights, exaggerated or large dance movements, camera push/pull, camera zoom, camera pan, camera tilt, complex hot-pot props, oversized eating motions, smile or expression while sleeping, passersby, silver panels, silver center console, gray/fabric seats, Oriental Pearl Tower, Lujiazui skyline, American-style empty distant landscape, subtitles, text, watermarks, CG feel, anime style, plastic skin.

[Mark H @markh_app]

天啊 这也太爽了吧 Cybercab自己开 我想干嘛就干嘛 😌
Only on Cybercab. Dance, dumplings, nap on a giant plushie.

Prompt 👇

[Mark H @markh_app]

Seeddance prompt 👇🏻

Real-life footage, shot on a phone. 16:9 landscape, real-time speed throughout — no slow motion at any point. Use the uploaded first-frame image as the starting frame: appearance, outfit, setting, and props follow that image exactly, no extra description needed — the video continues naturally from that frame. Only the female protagonist appears on screen; no other person or body part should appear. Her friend, who is filming, never appears on camera — only her handheld phone POV, with a natural, slightly shaky handheld feel.

Outfit (already shown in the first frame, for reference only):
White ribbed spaghetti-strap camisole, sweetheart neckline; ultra low-rise relaxed black lounge trousers with side-waist cutouts. Silver necklace, rings, smartwatch.

Camera:
Friend stands in the living room filming on her phone. 16:9, medium-wide shot, full body in frame, camera at roughly chest height, eye-level or slightly above — never low-angle toward the body's center. Handheld selfie-style texture: slight motivated shake, occasional autofocus hunting, a sense of the operator's breathing — no deliberate cinematic movement.

Pacing and action (real-time throughout):

0-2s: She's just sat on the kids' balance bike, legs naturally bent straddling it, feet lightly touching the floor. She looks down at the mini bike, smiling like she's thinking "can this thing even move?"

2-4s: She pushes off gently with both feet, slowly gliding around the room. The bike's too small for her, so her legs stay bent — a bit silly-looking but natural. She gets more into it, smile widening.

4-6s: She keeps gliding easily, absorbed in the fun, not looking at the camera, body swaying naturally with the bike.

6-8s: Curious, she reaches out her right index finger and lightly presses the center of the handlebar, like pressing an activation button.

8-11s (Hook, key motion shot): She lifts both feet off the ground, tucking them so they rest steadily on the frame bars either side of the front wheel — feet don't touch the floor and don't block the wheel's rotation. The bike starts moving forward on its own, with clear visible displacement — she and the bike glide together to a spot further into the room, wheels genuinely turning to drive it forward, not spinning in place. It shows self-balancing and self-steering: the frame subtly adjusts left-right balance, the front wheel completes a natural turn on its own, and the bike makes a small lane-change-like path shift, as if steering itself. She sways slightly from losing control, then breaks into a surprised, excited look, turns toward her friend off-camera, laughing: "Are you seeing this? This thing has FSD!" Off-screen, her friend's clear female voice gasps in surprise, then laughs. Friend never appears on camera.

11-13s: The bike keeps moving forward on its own, still turning and shifting path slightly, her feet still resting on the frame bars without touching ground. She looks down at the bike, then ahead, amazed and delighted.

13-15s: She gives a small amused head-shake, like experiencing this for the first time, shot ends naturally, not looking at the camera.

Image quality:
TikTok/Reels/Shorts-style phone footage — slightly soft, mildly blurry, visible noise/grain, phone HDR artifacts, low-bitrate compression, slight motion blur, minor auto-exposure shifts. No cinematic quality, no professional lighting, no excessive sharpness.

Prohibited:
No text, captions, or subtitles; no second person or other body parts in frame; the filmer must never appear on camera; composition must not emphasize chest, hips, or body curves.

[Mark H @markh_app]

离谱到家了，儿童平衡车都配上 FSD 了 😉
Even balance bikes have FSD now?? 💀

Prompt 👇

[Mark H @markh_app]

Seeddance 2.0 prompt:

Create a 15-second, 24fps, photorealistic 8K image-to-video. The upload is the exact first frame and sole visual reference. No cup in frame one.

【LOOK】Natural fixed in-car vlog: real skin, soft daylight, mild wide angle. No ad, studio, CGI, game-engine, or plastic AI look.

【WOMAN】Lock her identity, face, gray-green eyes, skin, proportions, high dark bun, bangs, and loose hair. Keep the same pale gray-white ribbed spaghetti-strap tank, two thin straps, V/U neckline, coverage, fit, and black belt. No necklace or body/wardrobe drift. Matcha adds pale-green stains and white foam; fabric stays ribbed and opaque.

【CAMERA — HIGHEST PRIORITY】Lock the exact first-frame camera, lens, framing, and angle for all 15 seconds. Show only her upper body, hands, belt, drink, and reactions. No pan, tilt, track, zoom, reframe, or angle change. Rigid mount; only tiny real chassis vibration.

One unbroken take: no cuts, inserts, second camera, windshield, road, or exterior view. She is the only visible person. No human figure, shadow, or reflection anywhere.

【CAR / FSD】Keep the same Cybertruck and interior. From frame one, window scenery moves backward with natural parallax/blur, then rapidly stops; never freeze, loop, reverse, or jump. No people outside. She stays alert, belted, looking forward. FSD drives; the untouched wheel makes small smooth corrections, then settles while braking. Never remove it or change the car.

【CUP】She holds nothing for 3 seconds. Around 3 seconds, her right hand retrieves a matcha latte from the unseen cupholder below frame. It enters continuously from the bottom. Without pausing, she brings it to her lips for a first sip near 4 seconds as braking begins.

Plain pale wide-mouth paper cup, fully open, showing green matcha, white foam, and hand-drawn sheep. No lid, seal, film, straw, sipping hole, logo, text, sticker, or QR code.

【OFF-SCREEN EVENT】A child runs into the road entirely off-camera, triggering FSD. Never show the child/event in windows, screens, mirrors, reflections, shadows, or blurred shapes. Never move/cut to explain it. No collision/injury; only her relief confirms safety.

【BRAKING / SPILL】Strong real emergency stop from city speed, neither gradual nor instant. Nose/suspension dips; scenery stops. Belt locks visibly across collarbone/chest while lap belt holds her pelvis. Torso, head, loose hair, and cup hand move forward naturally in sequence; belt catches her, then one small rebound. No impact, rigid/rag-doll motion, or repeated bouncing.

The open cup is near her lips. She keeps hold of it. As the belt catches her, her hand continues and wrist turns inward, aiming the rim toward her. Matcha/foam spill over the bare rim onto below her chin, top, right shoulder, arm, and lap. No lid exists. Noticeable, not explosive; sheep art breaks apart and stains remain continuous.

【TIMING】
0:00–0:03 — Cup-free; moving scenery; subtle wheel corrections; she watches forward.
0:03–0:04.2 — Retrieves open unmarked cup and brings it to her lips.
0:04.2–0:05.5 — Off-screen danger; eyes snap forward; FSD brakes hard; suspension dips, belt locks, body moves, wrist turns, matcha spills. Camera fixed.
0:05.5–0:10 — Stopped. She confirms safety, relaxes, exhales. Nobody else appears.
0:10–0:15 — She looks at the spill, then ruined sheep art: restrained disbelief. No speech, camera look, or laughter.

【AUDIO】Only real cabin/tire, braking, suspension, and belt-lock sound. Drink/spill is completely silent: no water, liquid, splash, canned, or comic effect. No voices, music, narration, scream, impact, or subtitles.

【AVOID】Second person/child visible; human reflections; cuts/camera motion; exterior/road shots; slow motion; background glitches; frozen/wild wheel; weak stop; loose belt; unnatural body motion; splash sound; idle cup holding; cup in first frame; any lid/straw/logo/text; dropped cup; collision/injury; transparent clothes; identity/wardrobe/anatomy drift; watermark, CGI, anime.

[Mark H @markh_app]

Cybertruck 自动驾驶 FSD 救了小孩，却救不了我的抹茶拿铁🍵😅

A child ran into the road. Cybertruck FSD stopped in time, but my matcha latte didn’t.

Prompt👇

[Mark H @markh_app]

Prompt:
# Cybercab - Delivery Rider Version

Generate a 15s, 24fps, 8K photoreal image-to-video. The uploaded first frame is the absolute reference for the woman, wardrobe, Cybercab, Shanghai setting, camera, light and composition.

【REFERENCE】Natural fixed in-car vlog; no selfie, ad, studio, 3D or AI look. Start exactly from frame one. Preserve her face/body, gray-green eyes, high dark bun, bangs and loose hairs. Keep the exact pale gray-white ribbed string-strap tank top: two ultra-thin straps, same low V/U neckline, opacity and fit. Black belt stays over it. No identity, wardrobe, strap, neckline or belt drift.

【FIXED CAMERA — HIGHEST PRIORITY】Lock frame-one angle, crop, lens and distance for all 15s. Camera is fixed low on front console, slightly right of center, aimed back-left, offscreen. Never turn, pan, track rider, zoom or reframe. Woman stays screen-left; empty seat screen-right; cabin/window stay aligned. No mirror, flip, cut or exterior shot.

The large side window already visible in frame one is the ONLY rider window, defined as Cybercab's LEFT window. Never show him on another side, create another window or turn camera toward him.

【CAR/SETTING】Preserve the same daylight, Shanghai street and traffic. Keep the exact two-seat driverless Cybercab and cabin. No wheel, yoke, pedals, instruments, shifter or extra passenger. Never Cybertruck.

【RIDER】Shanghai male delivery rider, 20–35, alone on a low-seat e-scooter with square delivery box. Yellow jacket/helmet; no readable logo. He stops parallel on Cybercab's LEFT, 60–90cm away, facing the SAME direction. Never oncoming or wrong-way.

He initially faces forward, unaware it is driverless. A peripheral glance notices the empty front cabin; he freezes, then quickly turns eyes and head. Eyes move first; head turns farther and leans slightly down/sideways. Brows rise subtly at no wheel or driver. His gaze scans inside, briefly notices the woman and their eyes meet. No improper intention or deliberate body gaze; awkwardness comes only from sudden eye contact.

【TIMELINE】

0:00–0:03 — Exact first frame. Cybercab stopped at red. Woman scrolls on phone low near lap. Rider arrives same direction and stops beside the existing left window, facing forward.

0:03–0:07 — Camera never moves. His peripheral glance detects the empty cabin; he freezes, turns eyes/head, then leans slightly down to confirm no wheel or driver. Woman still looks at phone.

0:07–0:11 — His gaze scans inside and briefly notices her. She looks up simultaneously. Their eyes meet about one second: ordinary awkwardness, not misconduct.

She uses two fingers to pinch the EXISTING pale ribbed fabric at her CHEST beside the black belt, lifts that tiny area once, then releases. Chest fabric, NOT neckline edge. Tiny instinctive motion. Same top, straps, neckline, coverage and belt remain unchanged; fabric returns immediately.

Rider stays neutral, neither smiling nor guilty. Eyes slowly return to road, then head turns forward. No nod, wave or second look.

0:11–0:15 — Cybercab remains STOPPED. Rider departs first in the same direction through the same window, never looking back. Woman releases fabric, returns hand to lap, and follows him with eyes plus a small head turn. Car/camera/cabin remain stationary to end.

【AUDIO】Quiet cabin, Shanghai traffic, scooter motor and city ambience. Both silent. No music, narration or captions.

【AVOID】camera move/track/reframe, cut, other window, wrong side/direction, passenger, missing delivery box, readable logo, head inside cabin, staring from start, skipping driverless surprise, prolonged body gaze, creepy/guilty look, flirting, smile/nod/wave, pulling neckline edge, redesigned neckline/top, strap drift, exaggerated covering, body deformation, Cybercab moving, face/cabin drift, wheel/yoke/Cybertruck, visible camera, speech, text, watermark, CGI/anime.

[Mark H @markh_app]

他只是在看 Cybercab 无人驾驶出租车……是我误会外卖小哥了 😅

He was just looking at the driverless Cybercab cruising in Shanghai..

Prompt 👇

[Mark H @markh_app]

在上海坐 Cybercab，结果尴尬了…… 😅
Riding a Cybercab in Shanghai got awkward…

Learning: The hardest part wasn’t the character—it was getting the vehicle’s scale relative to its surroundings physically accurate.

ChatGPT + Seedance 2.0 👇

[Mark H @markh_app]

昨天被 Starship 吓了一大跳。

From yesterday: Starship’s launch got scrubbed… but it still found a way to make me jump.

Best of luck to Starship on its next launch attempt in a few days. Really hoping everything goes smoothly this time. 🤞

ChatGPT + Seedance 2.0 👇

[Mark H @markh_app]

Seeddance 2.0 Prompt:

Create a 15s, 24fps, 8K photoreal video. @selfie controls woman, Cybercab, camera and composition. Natural fixed in-car vlog; mild-wide optics, 180° blur, real physics. No handheld, studio, 3D/game or AI look.

REFERENCE — Preserve @selfie’s exact face, eyes, skin/body, bun/bangs, ribbed string-strap top, belt, cabin, seats, armrest, materials, light, perspective and crop. No redesign/drift.

POSITION — She stays in @selfie’s occupied seat, defined as Cybercab’s RIGHT passenger seat. Woman stays on LEFT screen at identical size/height; empty seat RIGHT; window far RIGHT. Never swap/mirror/flip; belt diagonal unchanged.

ONE CAMERA — Device is fixed on low front console, right of center and below eye level, aimed backward, offscreen. One take: lock position, angle, lens, crop and distance; woman, seats, armrest, pillars and window stay aligned. No cuts, other angle, close-up, exterior, pan, zoom, shake or reframing.

LOOK — Daytime natural diffuse 6500K sky/window light: soft, even, low contrast, both cheeks equally lit. No hard/studio/AI light or waxy shine. Keep matte-black cabin, natural skin, pale top, black belt, blue sky; no red/silver trim. Real pores, fine hair, flush, imperfections and eye catchlights; no plastic/oily skin. Restrained micro-expression; no deep breath, sigh or posing. Subtle road vibration; camera stable.

CYBERCAB — Keep exact two-seat driverless Tesla Cybercab in @selfie. No wheel/yoke, column, pedals, driver display, shifter or cockpit. Never Cybertruck/sedan. No extra seat/passenger.

STARSHIP — A recognizable REAL SpaceX Starship, not sci-fi: VERTICAL upright, nose skyward; tall slender stainless cylinder, weld rings, black hexagonal heat-shield tiles on one side, two forward/two aft flaps. Base secured vertically to wheeled SPMT. It moves at walking speed on a parallel road 100–150m away. Window shows most of upright rocket and transporter. Never floats, lies horizontal or passes overhead.

STORY — Gold Cybercab drives autonomously on Texas Highway 4 toward Starbase. She stays belted in same right passenger seat, hands in lap, never driving.

0:00–0:04 — Continue from @selfie. South Texas plain, scrub, poles and distant launch structures slide past with natural blur. From first frame eyes move outside, then head turns slightly toward right window, quietly curious. No sigh or look to lens.

0:04–0:08 — No cut. Escort vehicles/equipment enter same window. Curiosity grows; gaze tracks outside, head turns farther, torso leans only slightly. Same seat, restrained motion.

0:08–0:12 — Starship passes upright at mid-distance on SPMT; nose, steel cylinder, black tiles, four flaps and wheels read clearly. Crews walk nearby; spectators stay behind barriers. No ignition/flame/smoke. Her eyes widen slightly, expression brightens, lips part subtly: quiet authentic excitement. Silent, no “Wow.”

0:12–0:15 — Starship continues past same window. She stays focused outside, never returns to lens. One hand lightly braces on seat edge/armrest; upper body lifts only slightly and leans windowward; neck extends gently forward/up while gaze follows rocket, eager to see more. Hips stay seated, belt fastened, no seat change/window touch. Hold last half-second watching outside.

AUDIO — EV hum, tires, distant machinery/crowd. Woman silent. No music/captions.

AVOID — mirror/flip, seat swap, woman screen-right, camera change, cabin drift, Cybertruck, wheel/yoke, extra passenger, visible device, arm to lens, miniature rocket, UFO/saucer, horizontal/floating rocket, building/canopy/bridge, wings, rocket near window/overhead, ignition/explosion, sigh, speech/“Wow,” looking at lens, leaving seat, scream, huge mouth/smile, text, watermark, CG/anime.

[Mark H @markh_app]

No steering wheel. No driver.

Just a Cybercab taking me to Starbase as Starship rolls out for tomorrow’s launch.

Had to see this up close.

ChatGPT + Seedance 2.0
Full prompt below 👇

[Mark H @markh_app]

@ForestChenfeng Sure! Here's the character reference image I used to test out SeedDance 2.0 Mini. The clothing, wardrobe, and accessories are all specified directly in the prompt.

[Mark H @markh_app]

Took myself on a night tour of Shanghai 🌃

Learning: Me almost looking too perfect 😅

ChatGPT + Seedance 2.0 mini

Prompt below 👇
SHANGHAI DARK-STREET SELFIE VLOG — 15 SEC, FAST MUSIC-CUT MONTAGE, VERTICAL 9:16

Entire video is SELFIE-STYLE in EVERY shot: she holds her phone at natural arm's length, front-camera POV, her face and upper body framed close as in a real handheld selfie video — NOT filmed by another person.

[CHARACTER LOCK — identical in EVERY shot]

Same young woman throughout; her face must stay IDENTICAL to the reference image in every shot — same facial features, bone structure, skin tone. Hair: short brown bob with wispy fringe/bangs, follow the reference image. Do not beautify, restyle, or let her face drift or blend with other people. WARDROBE (never changes, follow the reference image for all detail): dark streetwear — a white strapless tube top with a small cross print, a high-waisted fitted black mini skirt, bare legs, black pointed-toe high heels, an oversized glossy black leather jacket worn off the shoulders. NO choker or neck accessory. Slim natural build. She is always the sharp foreground subject; other people stay softly blurred behind her.

STYLE: 26mm phone front-camera look, deep focus, realistic skin texture, mild HDR, neutral iPhone color science, candid unedited feel, moody low-light night grading, no light trails, no fake motion blur. Fast hard cuts on the beat. Keep all expressions subtle and cool — small smiles, slight shifts, never exaggerated or theatrical.

(0:00–0:0215) 1933 Old Millfun at night — arm's-length selfie inside the dramatic concrete ramps and interlocking geometric architecture, dim moody uplighting, she turns slightly, faint cool expression.

(0:0215–0:043) Neon back-alley — selfie leaning against a wet concrete wall in a narrow alley washed in pink-and-blue neon signage, calm confident look. Night.

(0:043–0:0645) Elevated pedestrian bridge — selfie on a city skywalk at night, blurred traffic light streaks and towers behind her, breeze in her hair, relaxed cool gaze. Night.

(0:0645–0:086) teamLab digital light ★ SIGNATURE — selfie, she looks up slowly, blue/violet/gold projection washing over her face, face sharp, lights diffused. ~60% slow motion. Quiet awe.

(0:086–0:1075) Neon-lit metro station — selfie on an empty late-night platform, cool fluorescent and LED light, faint smirk. Night.

(0:1075–0:129) Exotic-pet reptile cafe ★ — selfie, a small chameleon perches on her shoulder; it flicks its tongue toward her face; she gives a small startled blink then a soft laugh. Cool indoor light, chameleon sharp.

(0:129–0:15) White Magnolia Skywalk finale — selfie on the newly opened North Bund high-altitude open-air observation deck, glittering Shanghai night skyline far below and across the river, wind moves her hair, quietly impressed. Night.

[Mark H @markh_app]

After what happened last night, I decided to learn how to fight.

ChatGPT + Seedance 2.0

[Mark H @markh_app]

Picked up a bow for the first time in years. Muscle memory is a strange thing.

ChatGPT + Seedance 2.0 Prompt below 👇

[IDENTITY LOCK — HIGHEST PRIORITY]

[ref-image] is the single source of truth for face and hair.

LOCK: eye shape, eyebrows, nose, lips, jaw and bone structure, skin tone, freckles, hairstyle (dark loose messy bun, soft bangs, loose face-framing strands).

DO NOT LOCK: head angle, face orientation, neck angle, gaze or expression. The reference is a frontal studio photo. Ignore its camera angle and expression. Match only identity. Render the face naturally for the body pose.

Rule: identity and hair follow [ref-image]. Angle, gaze and expression follow this prompt. No ethnicity labels or facial-feature descriptions.

[WARDROBE LOCK]

White ribbed crop tank, scoop neck, thick straps, hem just above the navel. Black washed low-waist denim mini skirt with raw frayed hem. Thin silver bracelet on LEFT wrist only. Black wrist release aid on RIGHT wrist. Bare ears. No earrings. Outfit stays identical for the full 15 seconds.

[GEOMETRY & FRAMING]

Right-handed archer. Left hand grips the bow, right hand draws. She stands in profile facing frame-right. Camera stays on her right side, slightly behind shoulder height. Right cheek and right arm are closest to camera. Bow remains in the right third of frame. Never mirror.

Shot size is locked for the entire take: waist to top of head with a little space above the bun. No zoom or push-in.

[EQUIPMENT]

Camo compound bow with cams, cables, sight and stabilizer. Carbon arrow with white and green fletching. Black wrist-strap release clipped to the D-loop. No finger tab or glove.

[EYES]

Both eyes stay open throughout. She breathes, opens her eyes wide, and locks onto the distant target. Only natural blinks under 0.25 seconds. Never closes one eye.

[DRAW & RELEASE]

At full draw, the release hand anchors beneath the jaw. Upper string touches the tip of her nose. Lower string presses lightly across the front of her chest against the tank top. These contacts remain until release. On release, the string snaps forward, vibrates, the bow rolls forward naturally, and the release hand moves back past her ear.

[REALITY]

Handheld mirrorless camera, 85mm equivalent, wide aperture.

Autofocus briefly catches the bow cables before returning to her face with visible focus stepping.

Wind blows from downrange toward camera (frame-right to frame-left), pushing bangs backward without covering her eyes. Hair strands and arrow fletching move naturally.

The operator handholds the camera with subtle breathing drift and makes a slight delayed flinch after the shot.

[ENVIRONMENT]

Outdoor archery range. Bright overcast daylight. Grass shooting lane, straw targets and paper faces downrange, soft tree line in the distance.

[SCENE — 15s, 16:9, ONE CONTINUOUS TAKE]

0-5s: At the shooting line, bow lowered. She clips the release onto the D-loop, settles her stance and exhales. Silent.

5-10s: She raises the bow and draws to full anchor. The operator walks a slow arc behind her right shoulder without changing shot size. She holds steady, eyes fixed on the target.

10-15s: Camera settles behind her right shoulder. THREE-QUARTER REAR VIEW showing the back and right side of her head, bun, right shoulder and upper back. Her face stays turned fully downrange, revealing only a slight edge of her right cheek and jaw. No frontal face. She releases, holds follow-through briefly, then lowers the bow.

[NEGATIVE]

No zoom or push-in. No mirrored or left-handed pose. No closed eyes or long blinks. No hair covering the eyes. No wind from behind. No earrings. No arrow flight or target impact. No selfie angle, drone, dialogue, subtitles, cuts, second person, operator shadow, extra limbs, broken bow geometry, string intersecting the body, beauty filters, skin smoothing, HDR, heavy grading, lens flare, logos, text, watermark or wardrobe changes.

[Mark H @markh_app]

"Quiet down!" — my neighbor, mid-drum-solo.

ChatGPT + Seedance 2.0
Prompt below 👇

STYLE

Live-action smartphone footage with authentic handheld realism. 16:9 landscape, real-time playback speed. The frame contains only the female lead @683d5152-fc45-48ce-baed-55df9f04ce65 from beginning to end. No other person may appear on screen at any time. She is playing a drum kit in her own living room.

SCENE & COMPOSITION

A warm, cozy corner of a home living room with natural ambient lighting. The lighting is intentionally imperfect and uneven, like a casual phone recording at home rather than a professional production. The main light comes from a single window on one side, leaving one side of her face brighter while the other falls into soft natural shadow.

She wears a black strapless dress with a deep V-neckline. The fabric is matte, soft, and naturally follows her slim figure with a clean, minimalist silhouette and almost no embellishments. The contrast between her skin, collarbones, and the black dress creates an elegant visual balance.

She has a naturally slim physique with no visible muscular definition.

She sits behind a full drum kit. The subject occupies the right side of the frame, while the drum kit extends toward the left and foreground. Medium close-up framing from the waist up.

The camera is positioned directly in front of the drum kit at approximately cymbal height, slightly below her face, looking through the gaps between the front cymbals and stands. Her upper body and face remain clearly visible throughout.

The camera remains completely stationary.

TIMING & ACTION (REAL-TIME ONLY)

0-2s

She sits quietly behind the drum kit in a relaxed posture. Her eyes are lowered toward the drums. She is calm and hasn't started playing yet.

2-3s

Without warning, she strikes the first powerful beat. Her entire body comes alive instantly, creating a dramatic shift in energy through physical performance rather than eye contact with the camera. The snare and hi-hat hit hard with explosive impact.

3-9s

She launches into an energetic, fast-paced drum performance with complete focus.

Every heavy beat drives her entire body. As each stroke lands, her shoulders and upper body naturally drop and rebound with the rhythm. Her head nods subtly on the beat, and her body rocks forward and backward with convincing weight and momentum. Every movement feels decisive, powerful, and physically grounded.

She remains fully immersed in the performance and never looks at the camera.

The drumming stays energetic and forceful throughout.

9-11s

While she's completely absorbed in playing, a series of muffled knocks suddenly comes from outside the wall:

"Boom... Boom... Boom..."

The sound is clearly coming from off-screen, muted by the wall.

Immediately afterward, a muffled male voice from outside the room says:

"Quiet down."

The voice remains completely off-screen, as though coming through the wall.

The neighbor must never appear on screen.

No face, body, shadow, reflection, or any other person is ever visible.

The female lead remains the only visible person throughout the entire video.

11-15s

She immediately stops playing.

Both hands freeze.

The drum sound cuts off abruptly.

She slowly turns her head toward the direction of the sound outside the frame.

She gives a restrained, subtle look of mild annoyance, consisting of nothing more than a small eye roll or a slight curl of one corner of her mouth.

The reaction feels authentic, understated, and effortless.

No exaggerated expressions, funny faces, nose scrunching, or overacting.

She ends the video in a relaxed, slightly dismissive mood.

The female lead never speaks at any point.

EXPRESSION

During the performance she is playful, confident, and completely focused on drumming. She never looks into the camera or poses for it.

At the end, her expression becomes subtly dismissive with a restrained eye roll or a slight smirk. The emotion should feel natural, understated, and believable, never exaggerated.

IMAGE QUALITY

Intentionally imitate an ordinary social-media smartphone recording rather than a polished production.

Use slightly soft focus, reduced sharpness, visible digital noise, smartphone filter characteristics, low-bitrate compression artifacts, and subtle motion blur.

Do not produce a 4K, ultra-sharp, cinematic, or perfectly lit image.

The framing should remain clean and visually balanced despite the casual smartphone aesthetic.

NEGATIVE PROMPT

No text, subtitles, captions, logos, or watermarks. No second person, no additional people, no reflections revealing another person, no shadows of another person, no hands entering from off-screen, no camera operator visible. The female lead must remain the only visible person throughout the entire video. No cinematic lighting, no studio lighting, no 4K crystal-clear look, no CGI appearance, no anime style, no game-engine rendering, and no slow motion.