《H3 多机位分屏实战：如何让两个画面真正属于同一个人、同一个动作、同一条时间线》
H3 分屏拍摄：一张图，就能做出双机位效果
最近在用 H3 做一些分屏视频，发现一个很好玩的用法：
只需要一张人物参考图，就可以让同一个人物同时出现在两个不同机位的画面里。
比如上面是正常的舞蹈主机位，下面则是一台直接放在地上的手机，从非常低的角度向上拍。
重点不是生成两个不同的表演，而是：
一个人、一次真实表演、同一个空间、同一条时间线，两台摄影机同时拍摄。
这样人物在上面做什么，下面的机位也必须同时看到对应的真实动作。
例如上面蹲下，下面就看到人物从地面机位前蹲下来；上面抬手，下面也同步看到抬手；人物向镜头靠近，两个机位都产生对应的空间变化。
H3 提示词模板
实际使用时，我一般只需要提供：
<Picture 1> = 人物参考图
然后用下面这种结构控制分屏：
这套模板的关键，其实就是把**“分屏”从后期效果变成现场多机位拍摄**。
舞蹈动作部分则可以继续结合我之前整理的舞蹈导演 Skill，把脚步、重心、躯干、肩、手臂、手腕、头部之间的动力链写进去。
如果想继续扩展玩法，可以参考之前的**“三格宫”分屏思路**：不一定非要固定三格，也可以改成上下双机位、大小画面、特殊低机位、侧面机位、近距离手机机位等。
所以这个方法最有意思的地方就是：
一张图 + 一个 Prompt，就可以让 H3 从“单机位生成”进一步变成“多机位拍摄”。 如果需要跟多扩展可以参考https://x.com/LoveUolanda/status/2092101996195311792 /// 提示词 subject_definitions:
<Subject 1> is the exact same single physical woman shown in <Picture 1>.
<Picture 1> is the sole and absolute visual authority for <Subject 1>'s identity and all existing visual assets. Fully preserve the exact referenced person, facial identity, facial structure, hairstyle, hair, body proportions, outfit, accessories, skin appearance, photographic texture, lighting character, and overall visual continuity from <Picture 1>.
Do not redesign, beautify, reinterpret, replace, or invent any appearance attribute of <Subject 1>.
Do not verbally redefine her appearance.
<Saxophone 1> is a realistic physical saxophone held and played by <Subject 1>. It is an additional physical prop required by the action, not a replacement for or modification of any existing visual asset in <Picture 1>.
There is ONLY ONE physical <Subject 1> and ONLY ONE physical saxophone in the shared scene.
summary:
[reference generation]
Create an original 15-second 9:16 vertical dual-camera music-performance composition.
<Subject 1> does NOT dance.
She remains focused on playing the saxophone throughout the performance.
The final frame is permanently divided into TWO fixed horizontal camera regions.
UPPER REGION = CAMERA A: A large normal-perspective master camera showing the primary saxophone performance.
LOWER REGION = CAMERA B: A completely independent smartphone camera physically placed directly on the floor, with its lens extremely close to ground level and pointing sharply upward toward <Subject 1>.
The two camera feeds are simultaneous views of ONE REAL PHYSICAL PERFORMANCE.
ONE WOMAN. ONE SAXOPHONE. ONE ROOM. ONE PERFORMANCE. ONE TIMELINE. TWO CAMERAS.
The woman is not performing separately for the two panels.
Every physical movement occurs exactly once and is simultaneously visible from both camera viewpoints.
retention_analysis:
<Subject 1>: fully_preserved.
The exact identity and appearance from <Picture 1> must remain unchanged throughout both camera feeds.
The lower camera does not generate another person.
Both camera feeds observe the same physical woman at the same physical location and the same moment.
The only major visual difference between the two regions is camera position, camera height, perspective, lens characteristics, and framing.
The saxophone is also the SAME physical saxophone in both camera feeds.
Do not generate two different saxophones.
Do not change the saxophone's shape, material, position, or orientation inconsistently between the two camera views.
shared_physical_world:
Both cameras are physically located inside the same room.
They operate simultaneously.
Camera A and Camera B observe the same performer from different physical locations.
The floor, walls, furniture, lighting, shadows, reflections, and background geometry remain spatially consistent.
The woman remains in one physically plausible location.
The saxophone remains attached to her hands and mouth through physically correct contact.
The mouthpiece remains at her mouth while she plays.
Her hands remain correctly positioned on the saxophone keys.
No duplicated body.
No duplicated saxophone.
No contradictory hand positions.
No teleportation.
No alternate performance.
camera_system:
CAMERA A — MAIN SAXOPHONE CAMERA:
The upper region occupies approximately 72% of the 9:16 frame.
Camera A is a stable normal-perspective performance camera positioned around eye level or slightly below eye level.
It faces <Subject 1> from a frontal or subtle three-quarter angle.
The camera remains mostly locked-off.
The composition prioritizes the woman's face, upper body, hands, and saxophone.
The saxophone must remain clearly readable.
The framing should feel like a professionally filmed intimate live music performance.
The camera does not pull far away merely to reveal the entire body.
CAMERA B — FLOOR-PLACED SMARTPHONE:
The lower region occupies approximately 28% of the total frame.
Camera B is a real smartphone physically placed directly on the floor.
The smartphone rests on the ground.
The lens is approximately 5–15 cm above floor level.
The lens points sharply upward toward <Subject 1>.
This is an EXTREME LOW-ANGLE FLOOR SHOT.
It must look as though someone literally placed a smartphone on the floor and started recording upward.
It is NOT a normal cinema camera tilted upward.
It is NOT a waist-level low-angle shot.
It is NOT a crop of Camera A.
It is NOT a digitally zoomed version of Camera A.
It is NOT a second independent scene.
The floor must feel physically close to the smartphone lens.
The lower image should contain strong floor-level perspective and visible upward spatial geometry.
The woman should visually rise above the low camera.
The face, saxophone, arms, and upper body can become dramatically foreshortened from below while remaining physically accurate.
The camera remains completely fixed on the floor.
No operator movement.
No tracking.
No orbit.
No crane movement.
No floating camera.
No digital zoom.
master_performance:
The physical performance must be designed FIRST as one continuous saxophone performance.
The two cameras only observe this single performance.
<Subject 1> stands naturally and plays the saxophone.
Her performance is expressive but controlled.
She uses realistic saxophone-playing posture:
stable stance, natural breathing, mouthpiece contact, both hands controlling the keys, subtle finger movements, small wrist adjustments, gentle torso movement, natural shoulder motion, subtle head movement, controlled swaying caused by musical phrasing.
She does NOT perform dance choreography.
No exaggerated dancing.
No jumping.
No large theatrical choreography.
No random arm waving.
The physical behavior must remain believable for a woman actually playing a saxophone.
detailed_description:
[Shot 1]
At 00:00.000, establish the permanent two-camera composition.
The upper region contains Camera A.
The lower region contains Camera B.
<Subject 1> is physically standing in the same room and holding <Saxophone 1>.
Both cameras see her simultaneously.
Camera A presents the normal intimate music-performance view.
Camera B presents the extreme floor-level smartphone perspective.
The woman brings the saxophone naturally into playing position.
The mouthpiece reaches her mouth.
Her fingers settle naturally over the keys.
Both camera feeds must show the same physical preparation.
[Shot 2]
From approximately 00:00.000–00:03.000, <Subject 1> begins playing the saxophone.
Camera A shows her face, upper body, hands, and instrument clearly.
Camera B simultaneously shows the exact same performance from the floor.
Her fingers move over the saxophone keys in both views at exactly the same physical moments.
Her breathing and embouchure remain consistent.
If her head tilts slightly in Camera A, Camera B must show that same head movement from below.
If her shoulders move with the musical phrase, the same shoulder movement must occur in both views.
The lower camera does not invent a separate performance.
[Shot 3]
From approximately 00:03.000–00:06.000, the saxophone performance becomes slightly more expressive.
<Subject 1> subtly shifts her weight from one leg to the other.
She slightly leans forward while continuing to play.
This is one physical movement.
Camera A records the subtle forward lean normally.
Camera B records the same physical lean from below.
As she leans forward, her upper body and saxophone physically move closer to the floor-level smartphone.
The lower camera should therefore produce a naturally stronger perspective change.
Her face and saxophone may become larger within Camera B because the physical distance between performer and smartphone decreases.
This enlargement must come from real spatial perspective, not digital zoom.
[Shot 4]
From approximately 00:06.000–00:09.000, <Subject 1> returns naturally toward her original upright posture while continuing to play.
Her fingers continue realistic saxophone-key movements.
She performs a small expressive head movement synchronized with the musical phrase.
Camera A sees the movement clearly.
Camera B sees the same movement from underneath.
The saxophone remains continuously connected to her mouth and hands.
No discontinuity.
No object duplication.
No hand deformation.
No floating instrument.
[Shot 5]
From approximately 00:09.000–00:12.000, the performance enters a more emotional musical phrase.
<Subject 1> slightly lowers her upper body and brings the saxophone subtly forward.
Her eyes and facial expression remain focused on the performance.
Her hands continue coordinated fingering.
The physical distance between her and the floor-level smartphone becomes slightly smaller.
Camera B emphasizes this dramatically.
The saxophone may partially dominate the lower-camera foreground because the instrument physically approaches the low lens.
The camera itself remains completely stationary.
Camera A continues to show the full relationship between her face, body, and saxophone.
Both feeds remain synchronized.
[Shot 6]
From 00:12.000–00:15.000, the saxophone performance reaches its emotional climax.
<Subject 1> performs one final expressive musical phrase.
Her torso makes a subtle natural rise.
Her shoulders lift slightly with the musical intensity.
Her head makes a small controlled movement.
Her fingers perform a final coordinated sequence on the saxophone keys.
At approximately 00:13.000, she subtly leans toward the camera while continuing to play.
Camera A captures this as an intimate performance gesture.
Camera B captures the same physical lean from the floor.
Because Camera B is only a few centimeters above the floor, the saxophone and upper body become dramatically larger and more imposing from below.
The perspective must feel physically authentic.
At the final musical note, she holds the saxophone naturally in playing position for a brief moment.
Both cameras hold the same final physical state.
CAMERA B VISUAL LANGUAGE:
The lower shot must strongly resemble an actual smartphone placed on the floor.
The lens is almost at ground level.
The floor is physically close to the lens.
The performer towers above the camera.
The camera points strongly upward.
The body and saxophone extend vertically through the frame.
The perspective is naturally wide and slightly exaggerated like a smartphone wide-angle lens.
The ceiling or upper background may appear behind the performer because of the upward camera direction.
The saxophone may visually project toward the lens when the performer leans forward.
The image should feel like an experimental behind-the-scenes music-video shot captured by a phone resting on the ground.
It should NOT feel like an expensive cinema camera placed on a low tripod.
It should NOT feel like a normal portrait camera tilted upward.
It should NOT feel like a crop of the upper panel.
SAXOPHONE PHYSICAL CONTINUITY:
<Saxophone 1> is one physical object.
The mouthpiece remains correctly positioned at <Subject 1>'s mouth.
The hands remain correctly positioned on the instrument.
Finger movements correspond naturally to the keys.
The saxophone strap, if physically required by the instrument and performance, remains consistent.
No duplicate instruments.
No disappearing instrument.
No instrument changing shape.
No impossible finger placement.
No floating saxophone.
No instrument passing through the face or hands.
SYNCHRONIZATION RULES:
ONE PHYSICAL WOMAN.
ONE PHYSICAL SAXOPHONE.
ONE PHYSICAL PERFORMANCE.
ONE SHARED ROOM.
ONE SHARED TIMELINE.
TWO SIMULTANEOUS CAMERAS.
Every action happens once.
Both cameras observe that exact action simultaneously.
Same breathing movement.
Same finger movement.
Same head movement.
Same shoulder movement.
Same torso movement.
Same lean.
Same final pose.
Different perspective only.
The camera difference is SPATIAL, never TEMPORAL.
Do not allow Camera B to improvise.
Do not allow Camera B to freeze while Camera A performs.
Do not allow Camera B to perform a different musical gesture.
Do not allow the performer to have contradictory poses.
COMPOSITION LOCK:
Permanent 9:16 vertical composition.
Upper region approximately 72%.
Lower region approximately 28%.
One thin fixed horizontal boundary.
The boundary never moves.
The upper region never expands fullscreen.
The lower region never expands fullscreen.
No cuts.
No wipes.
No transitions.
No camera-feed swapping.
No third camera.
No additional performer.
No duplicated performer.
No reflection creating another performer.
No teleportation.
No disappearing body parts.
No asynchronous movement.
No separate choreography.
IDENTITY LOCK:
<Picture 1> is the sole visual authority for <Subject 1>.
Do not redesign the person.
Do not beautify the face.
Do not alter facial proportions.
Do not enlarge the eyes.
Do not change hairstyle.
Do not change hair color.
Do not add makeup.
Do not change clothing.
Do not add accessories.
Do not reinterpret the character.
The low-angle perspective may naturally alter apparent spatial proportions, but must never alter the actual identity or visual design established by <Picture 1>.
overall_soundscape:
One shared physical room ambience.
The same saxophone performance is heard from the same physical event.
Natural breath and subtle instrument-key sounds remain synchronized with the visible performance.
Subtle room reverberation.
Natural saxophone resonance.
No dialogue.
No spoken words.
No singing.
No additional instruments unless musically required by the background track.
non_diegetic_music:
A cinematic intimate instrumental backing track designed around a live saxophone performance.
Warm atmospheric pads, restrained piano or electric piano, subtle percussion, gentle bass, spacious harmonic texture, emotional but elegant pacing.
The saxophone is the primary foreground musical element.
The music must support a believable live saxophone performance rather than overpower it.
Musical phrasing should correspond naturally to the performer's breathing, finger movements, head movement, subtle torso movement, and final sustained note.
The entire audio structure represents ONE continuous saxophone performance observed simultaneously by TWO cameras.