像当初见面那样

分镜与 MiniMax H3 提示词 · 按 01 → 05 顺序生成

每段依次上传女主参考、男主参考、本段分镜三张图，再复制该段提示词。色彩沿用原图，未追加调色。这里交付的是图片和提示词，尚未生成视频。

相邻画面与原图对照

左侧固定原图四；右侧可依序切换。比较时留意肤色、衣料颜色、背景、视线和人物位置。

原图四 — 原图四 · 对照基准

待对照分镜 — 第二镜第 1 段 · 男主对白

2 / 5

第一镜 · 女主问句 6 秒

第一镜 · 女主问句

我们该怎么告别呢？

依次上传： ① 女主参考 ② 男主参考 ③ 本段分镜

下载提示词 TXT

展开提示词

subject_definitions:
<Subject 1> is the same adult woman from <Picture 1>: the exact referenced face, eyes, long dark hair arranged with the original flower-and-pearl ornaments, dangling earrings, cream patterned robe and restrained red embroidered borders. <Picture 1> is an identity and costume reference, not the scene background.
<Subject 2> is the same adult man from <Picture 2>: the exact referenced face, long dark hair and topknot with the original pin, dark finely patterned robe, subdued collar borders and pale inner collar. <Picture 2> is an identity and costume reference, not the scene background.
<Picture 3> is the composition, starting-pose, scene, lighting and color anchor for [Shot 1].

summary:
[reference generation] A 6-second single-shot farewell question using <Picture 3> as the opening composition anchor. <Subject 1> quietly asks <Subject 2> how they should say goodbye, retaining both original identities. Only the woman speaks.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the reference identity, facial anatomy, hair arrangement, ornaments and costume; allow only the specified performance.
<Subject 2> (appears in [Shot 1]): fully_preserved - retain the reference identity, facial anatomy, hair arrangement and costume; allow only the specified performance.
<Picture 3> ([Shot 1] composition and starting-pose anchor): fully_preserved - retain its scene layout, initial positions, framing, existing illumination and color relationship. The specified later movement develops from this anchor.

detailed_description:
The target uses the same visual treatment as the supplied scene still. Retain the source still's exact existing exposure, white balance, skin and fabric colors, saturation, contrast, shadow direction and background structure. Keep these stable over time; do not add warm lighting, orange grading, a filter, bloom, flare, haze or an exposure transition. Preserve realistic skin and fabric texture, the established woman-left/man-right geography and original close spacing. Use a locked camera, only subtle breathing and minute natural hair movement. No extra person, prop, embrace, subtitles or on-screen text.
[Shot 1] Reconstruct the over-the-man's-shoulder close-up in <Picture 3>. <Subject 1> occupies the left and center, looking into the man's face; <Subject 2>'s dark back and shoulder occupy the right foreground. Preserve the source terrain and its existing illumination. Her mouth begins at rest, her eyes slightly moist, without a new stream of tears.
From the opening to approximately 00:00.600, she takes one small, unsteady breath while maintaining eye contact. From approximately 00:00.600 to 00:04.300, <Subject 1> (S1) says, <d>[Chinese] 我们该怎么告别呢？</d> Use a natural young-adult female Mandarin speaking voice in a normal mid register: clear, gentle and rounded, with only a little natural breath. Keep the same underlying timbre, habitual pitch range, accent and articulation throughout the sequence. Emotion changes the breath, pauses and slight tremor, not the vocal identity. She is speaking to the man beside her, not narrating to an audience. Her delivery is audibly choked with reluctance, held inside a soft conversational volume. Give the opening two syllables a small hesitation, carry the middle phrase on one breath, and let the closing question remain vulnerable rather than bright. Do not insert long gaps between individual characters or turn the line into a sob.
After the final syllable, her lips settle naturally. Leave a short reaction and a clean editing handle through 00:06.000. Her attention begins to drift upward only after the question has finished, providing motivation for the following upward-looking shot; no second sentence is added. The man stays silent and does not mouth a reply. Finish with the same exposure and ambient level as the opening.

overall_soundscape:
A very faint, steady outdoor breeze beneath close, intelligible dialogue, with natural breathing and minimal fabric rustle. Maintain the same low ambience level across the sequence. No additional voices, vocal improvisations or conspicuous sound effects.

non_diegetic_music:
N/A

第二镜第 1 段 · 男主对白 12 秒

第二镜第 1 段 · 男主对白

像当初见面那样！还是有点舍不得。真的有点舍不得。

依次上传： ① 女主参考 ② 男主参考 ③ 本段分镜

下载提示词 TXT

展开提示词

subject_definitions:
<Subject 1> is the same adult woman from <Picture 1>: the exact referenced face, eyes, long dark hair arranged with the original flower-and-pearl ornaments, dangling earrings, cream patterned robe and restrained red embroidered borders. <Picture 1> is an identity and costume reference, not the scene background.
<Subject 2> is the same adult man from <Picture 2>: the exact referenced face, long dark hair and topknot with the original pin, dark finely patterned robe, subdued collar borders and pale inner collar. <Picture 2> is an identity and costume reference, not the scene background.
<Picture 3> is the composition, starting-pose, scene, lighting and color anchor for [Shot 1].

summary:
[reference generation] A 12-second single-shot continuation of the farewell scene in the composition of <Picture 3>. <Subject 1> remains looking upward on the left while <Subject 2>, on the right, replies with three restrained lines. Only the man speaks.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the reference identity, facial anatomy, hair arrangement, ornaments and costume; allow only the specified performance.
<Subject 2> (appears in [Shot 1]): fully_preserved - retain the reference identity, facial anatomy, hair arrangement and costume; allow only the specified performance.
<Picture 3> ([Shot 1] composition and starting-pose anchor): fully_preserved - retain its scene layout, initial positions, framing, existing illumination and color relationship. The specified later movement develops from this anchor.

detailed_description:
The target uses the same visual treatment as the supplied scene still. Retain the source still's exact existing exposure, white balance, skin and fabric colors, saturation, contrast, shadow direction and background structure. Keep these stable over time; do not add warm lighting, orange grading, a filter, bloom, flare, haze or an exposure transition. Preserve realistic skin and fabric texture, the established woman-left/man-right geography and original close spacing. Use a locked camera, only subtle breathing and minute natural hair movement. No extra person, prop, embrace, subtitles or on-screen text.
[Shot 1] Open in the exact two-person close composition and starting poses of <Picture 3>. <Subject 1> is on the left, looking upward; <Subject 2> is to her right and slightly behind, looking at her face. His lips begin at rest. She remains visibly present, silent and attentive without turning toward him.
After a small opening breath, <Subject 2> (S1) speaks the following three lines from approximately 00:00.300 to 00:11.400. Use a natural young-adult male Mandarin voice in a low-to-mid register, low, warm in vocal timbre, resonant but restrained, with clear ordinary diction. Keep pitch, timbre, accent and articulation stable. Sadness remains suppressed in breath and word weight; do not manufacture a gravelly bass or raise the volume into a declaration.
He first says, <d>[Chinese] 像当初见面那样！</d> Let this run as one gentle phrase, with tenderness in the reference to their first meeting and a soft falling finish. The exclamation mark signals conviction, not a shout. Take about 2.8 seconds for the phrase, then pause about 0.7 seconds.
He continues, <d>[Chinese] 还是有点舍不得。</d> Allow a small hesitation of about 0.25 seconds after the opening two syllables, then keep the rest connected. The understatement is quiet and the final three syllables slow slightly without becoming separate isolated words. Allow approximately 3 seconds of speech, then about 0.85 seconds of silence.
He finally says, <d>[Chinese] 真的有点舍不得。</d> Give the opening two syllables firmer articulation, followed by a restrained pause of about 0.35 seconds. Complete the rest in one connected breath, with a slight caught quality at the finish. Allow approximately 3.2 seconds of speech. Maintain low volume and the same voice identity.
Through 00:12.000, his mouth returns to rest while he keeps looking at her. The woman's upward head angle and gaze remain continuous. Do not add an object word to either statement of reluctance, repeat a phrase, overlap speakers or start the woman's dialogue in this clip.

overall_soundscape:
A very faint, steady outdoor breeze beneath close, intelligible dialogue, with natural breathing and minimal fabric rustle. Maintain the same low ambience level across the sequence. No additional voices, vocal improvisations or conspicuous sound effects.

non_diegetic_music:
N/A

第二镜第 2 段 · 女主仰望近景 14 秒

第二镜第 2 段 · 女主仰望近景

他们都说~~~，忘记就不会痛苦。可时间并不会赋予任何一个人，忘记的权利……

依次上传： ① 女主参考 ② 男主参考 ③ 本段分镜

下载提示词 TXT

展开提示词

subject_definitions:
<Subject 1> is the same adult woman from <Picture 1>: the exact referenced face, eyes, long dark hair arranged with the original flower-and-pearl ornaments, dangling earrings, cream patterned robe and restrained red embroidered borders. <Picture 1> is an identity and costume reference, not the scene background.
<Subject 2> is the same adult man from <Picture 2>: the exact referenced face, long dark hair and topknot with the original pin, dark finely patterned robe, subdued collar borders and pale inner collar. <Picture 2> is an identity and costume reference, not the scene background.
<Picture 3> is the composition, starting-pose, scene, lighting and color anchor for [Shot 1].

summary:
[reference generation] A 14-second single-shot close-up in the composition of <Picture 3>. <Subject 1> begins her upward-looking response, suppressing tears while speaking clearly. <Subject 2>'s cropped shoulder remains on the right, and only the woman speaks.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the reference identity, facial anatomy, hair arrangement, ornaments and costume; allow only the specified performance.
<Subject 2> (appears in [Shot 1]): fully_preserved - retain the reference identity, facial anatomy, hair arrangement and costume; allow only the specified performance.
<Picture 3> ([Shot 1] composition and starting-pose anchor): fully_preserved - retain its scene layout, initial positions, framing, existing illumination and color relationship. The specified later movement develops from this anchor.

detailed_description:
The target uses the same visual treatment as the supplied scene still. Retain the source still's exact existing exposure, white balance, skin and fabric colors, saturation, contrast, shadow direction and background structure. Keep these stable over time; do not add warm lighting, orange grading, a filter, bloom, flare, haze or an exposure transition. Preserve realistic skin and fabric texture, the established woman-left/man-right geography and original close spacing. Use a locked camera, only subtle breathing and minute natural hair movement. No extra person, prop, embrace, subtitles or on-screen text.
[Shot 1] Match the woman's close-up and the small cropped male shoulder at the right edge in <Picture 3>. <Subject 1> keeps her chin lifted and eyes directed upward, exactly as she was before this cut. The framing changes, not her position in the scene. Her lips begin naturally at rest, with mild moisture in the eyes and no newly invented running tears.
From approximately 00:00.250, <Subject 1> (S1) begins, <d>[Chinese] 他们都说，忘记就不会痛苦。</d> Use a natural young-adult female Mandarin speaking voice in a normal mid register: clear, gentle and rounded, with only a little natural breath. Keep the same underlying timbre, habitual pitch range, accent and articulation throughout the sequence. Emotion changes the breath, pauses and slight tremor, not the vocal identity. She is speaking to the man beside her, not narrating to an audience. She starts softly and with uncertainty, as if trying to believe a familiar consolation. Take approximately 1.8 seconds for the first four syllables, slightly sustaining the last of those syllables. Pause about 0.5 seconds, then deliver the rest in approximately 2.5 seconds, connected and with the ending falling gently. Pause about 0.6 seconds.
Without changing her gaze or vocal identity, <Subject 1> (S1) continues, <d>[Chinese] 可时间并不会赋予任何一个人，忘记的权利……</d> The opening contrast gains quiet firmness rather than anger. Keep the phrase before the comma connected at an unhurried conversational rate, approximately 4.6 seconds, followed by a short 0.3-second pause. Speak the final five syllables in approximately 2.1 seconds, slightly slower and with a small catch in the breath. Complete the entire utterance by roughly 00:12.800, then allow approximately 0.8 seconds of natural breath and a small editing handle through 00:14.000.
The man stays silent and his shoulder remains stable. The woman does not turn her head or lower her eyes in this clip. Preserve the original collar and jewelry geometry as she breathes. Avoid extra pauses between each word, theatrical crying, a recitation rhythm or a new narrator voice.

overall_soundscape:
A very faint, steady outdoor breeze beneath close, intelligible dialogue, with natural breathing and minimal fabric rustle. Maintain the same low ambience level across the sequence. No additional voices, vocal improvisations or conspicuous sound effects.

non_diegetic_music:
N/A

第二镜第 3 段 · 双人稍宽仰望 15 秒

第二镜第 3 段 · 双人稍宽仰望

那些所有的回忆，都将永远存在于你我的脑海里，直到再次提起，直到再次想起，

依次上传： ① 女主参考 ② 男主参考 ③ 本段分镜

下载提示词 TXT

展开提示词

subject_definitions:
<Subject 1> is the same adult woman from <Picture 1>: the exact referenced face, eyes, long dark hair arranged with the original flower-and-pearl ornaments, dangling earrings, cream patterned robe and restrained red embroidered borders. <Picture 1> is an identity and costume reference, not the scene background.
<Subject 2> is the same adult man from <Picture 2>: the exact referenced face, long dark hair and topknot with the original pin, dark finely patterned robe, subdued collar borders and pale inner collar. <Picture 2> is an identity and costume reference, not the scene background.
<Picture 3> is the composition, starting-pose, scene, lighting and color anchor for [Shot 1].

summary:
[reference generation] A 15-second single-shot, modestly wider two-person view anchored by <Picture 3>. <Subject 1> continues the same upward-looking monologue while <Subject 2> listens quietly on her right. Only the woman speaks.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the reference identity, facial anatomy, hair arrangement, ornaments and costume; allow only the specified performance.
<Subject 2> (appears in [Shot 1]): fully_preserved - retain the reference identity, facial anatomy, hair arrangement and costume; allow only the specified performance.
<Picture 3> ([Shot 1] composition and starting-pose anchor): fully_preserved - retain its scene layout, initial positions, framing, existing illumination and color relationship. The specified later movement develops from this anchor.

detailed_description:
The target uses the same visual treatment as the supplied scene still. Retain the source still's exact existing exposure, white balance, skin and fabric colors, saturation, contrast, shadow direction and background structure. Keep these stable over time; do not add warm lighting, orange grading, a filter, bloom, flare, haze or an exposure transition. Preserve realistic skin and fabric texture, the established woman-left/man-right geography and original close spacing. Use a locked camera, only subtle breathing and minute natural hair movement. No extra person, prop, embrace, subtitles or on-screen text.
[Shot 1] Match <Picture 3>'s wider two-person framing. <Subject 1> remains on the left foreground with precisely the established lifted head and upward gaze; <Subject 2> remains on the right, slightly behind, looking at her. This is the next phrase of the same conversation, not a reset of either character's emotion or body position. Preserve the supplied visible robe boundaries without adding a belt, accessory or hand gesture.
Beginning after a small 0.2-second editing handle, <Subject 1> (S1) says, <d>[Chinese] 那些所有的回忆，</d> Use a natural young-adult female Mandarin speaking voice in a normal mid register: clear, gentle and rounded, with only a little natural breath. Keep the same underlying timbre, habitual pitch range, accent and articulation throughout the sequence. Emotion changes the breath, pauses and slight tremor, not the vocal identity. She is speaking to the man beside her, not narrating to an audience. Her sadness becomes tender as she speaks about shared memories. Allow approximately 2.8 seconds and a brief 0.3-second breath.
She continues immediately with the same speaker identity, <Subject 1> (S1), <d>[Chinese] 都将永远存在于你我的脑海里，</d> Use approximately 4.6 seconds, gently giving weight to the idea of permanence and to the two people sharing those memories. Keep the whole phrase connected rather than splitting every word. Allow a natural pause of approximately 0.5 seconds.
<Subject 1> (S1) continues, <d>[Chinese] 直到再次提起，</d> Take approximately 2.4 seconds and pause approximately 0.4 seconds.
<Subject 1> (S1) then says, <d>[Chinese] 直到再次想起，</d> Take approximately 2.5 seconds, fractionally slower and more emotionally exposed than the preceding phrase. Keep the two repeated constructions conversational rather than rhythmically recited.
All speech is complete around 00:13.700. Let a roughly 0.6-second breath and a short clean handle run through 00:15.000. Her eyes remain upward even at the end, so the next closer shot can continue the same physical state. The man is quietly engaged but never speaks or makes a new gesture. Keep the camera, source color, light, ambience and costume texture stable.

overall_soundscape:
A very faint, steady outdoor breeze beneath close, intelligible dialogue, with natural breathing and minimal fabric rustle. Maintain the same low ambience level across the sequence. No additional voices, vocal improvisations or conspicuous sound effects.

non_diegetic_music:
N/A

第二镜第 4 段 · 转头、对视、告白 8 秒

第二镜第 4 段 · 转头、对视、告白

同样你也要记得..... 我爱你。

依次上传： ① 女主参考 ② 男主参考 ③ 本段分镜

下载提示词 TXT

展开提示词

subject_definitions:
<Subject 1> is the same adult woman from <Picture 1>: the exact referenced face, eyes, long dark hair arranged with the original flower-and-pearl ornaments, dangling earrings, cream patterned robe and restrained red embroidered borders. <Picture 1> is an identity and costume reference, not the scene background.
<Subject 2> is the same adult man from <Picture 2>: the exact referenced face, long dark hair and topknot with the original pin, dark finely patterned robe, subdued collar borders and pale inner collar. <Picture 2> is an identity and costume reference, not the scene background.
<Picture 3> is the composition, starting-pose, scene, lighting and color anchor for [Shot 1].

summary:
[reference generation] An 8-second single-shot close two-person ending anchored by <Picture 3>. <Subject 1> first continues looking upward, then turns toward <Subject 2>, establishes mutual eye contact, and only then speaks her final confession. Only the woman speaks.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - retain the reference identity, facial anatomy, hair arrangement, ornaments and costume; allow only the specified performance.
<Subject 2> (appears in [Shot 1]): fully_preserved - retain the reference identity, facial anatomy, hair arrangement and costume; allow only the specified performance.
<Picture 3> ([Shot 1] composition and starting-pose anchor): fully_preserved - retain its scene layout, initial positions, framing, existing illumination and color relationship. The specified later movement develops from this anchor.

detailed_description:
The target uses the same visual treatment as the supplied scene still. Retain the source still's exact existing exposure, white balance, skin and fabric colors, saturation, contrast, shadow direction and background structure. Keep these stable over time; do not add warm lighting, orange grading, a filter, bloom, flare, haze or an exposure transition. Preserve realistic skin and fabric texture, the established woman-left/man-right geography and original close spacing. Use a locked camera, only subtle breathing and minute natural hair movement. No extra person, prop, embrace, subtitles or on-screen text.
[Shot 1] Begin in the close two-face framing of <Picture 3>. Both complete faces remain readable. <Subject 1> is on the left, still looking upward with the same head angle as before the cut. <Subject 2> is on the right and slightly behind, already looking at her face. Keep their torso positions and interpersonal spacing fixed. Use a natural young-adult female Mandarin speaking voice in a normal mid register: clear, gentle and rounded, with only a little natural breath. Keep the same underlying timbre, habitual pitch range, accent and articulation throughout the sequence. Emotion changes the breath, pauses and slight tremor, not the vocal identity. She is speaking to the man beside her, not narrating to an audience.
From approximately 00:00.300 to 00:03.100, <Subject 1> (S1) says, <d>[Chinese] 同样你也要记得。</d> Her tone is soft, serious and personal, gently asking him to remember. Keep her gaze UPWARD for this entire phrase. Deliver the words in one natural breath and let the last two syllables settle clearly. 
From approximately 00:03.100 to 00:04.300, she falls silent, lowers her upward gaze and slowly turns her head toward the man on her right. Her body stays still. He continues watching her and makes only the minute eye adjustment necessary to meet her gaze. Finish the turn and establish clear MUTUAL EYE CONTACT before the next line begins. This must be a continuous natural head movement, without a morph, jump, camera cut, embrace or added touch.
From approximately 00:04.400 to 00:06.400, while looking directly at his face and into his eyes, <Subject 1> (S1) says, <d>[Chinese] 我爱你。</d> Give the complete three-syllable phrase a full two-second speaking window in one connected breath. It is tender, definite and slightly choked, not a whisper that loses consonants. The final syllable must be fully voiced; do not cut it off, stretch each character separately or begin the line while she is still looking upward.
From approximately 00:06.400 through 00:08.000, hold the mutual gaze and allow a quiet, wordless breath after the completed confession. The man's mouth stays closed. Preserve the original face shapes throughout the turn, keep earrings attached naturally, and maintain stable illumination and source color through the final frame.

overall_soundscape:
A very faint, steady outdoor breeze beneath close, intelligible dialogue, with natural breathing and minimal fabric rustle. Maintain the same low ambience level across the sequence. No additional voices, vocal improvisations or conspicuous sound effects.

non_diegetic_music:
N/A