好用的 H3 Max 提示词更像一份简短的镜头说明:画面里有什么、发生什么动作、摄影机怎样观察、画面是什么风格,以及观众会听到什么。本页也覆盖 MiniMax H3 Max 提示词、H3Max Prompt 和 Max H3 提示词等常见搜索写法。
不要照抄所有形容词。选择一个案例,然后替换主体、动作、场景和声音,通常更容易得到稳定结果。
一条清晰的提示词可以拆成六个部分:
| 部分 | 应该写什么 | 示例 |
|---|---|---|
| 主体 | 人物、商品、动物或地点 | 一条漂浮在海滨城市上空的透明蓝鲸 |
| 动作 | 看得见的运动及变化 | 穿过高楼,然后转向日出方向 |
| 镜头 | 景别、机位与摄影机运动 | 低机位跟拍,随后升起并环绕主体 |
| 视觉 | 媒介、光线、颜色、材质和情绪 | 电影写实、金色薄雾、冷色玻璃反射 |
| 音频 | 对白、环境声、音效或音乐 | 风声、远处车流和低沉鲸鸣 |
| 时间线 | 复杂镜头的事件顺序 | 开始用近景,第 3 秒揭示城市,结尾停住 |
最短且实用的公式是:
主体 + 动作 + 镜头 + 光线/风格 + 音频。
只有在动作顺序确实重要时再加入时间线。5 秒视频最好只安排一个清楚动作;10–15 秒可以包含开始、发展和结尾。
At sunrise, a colossal translucent sky whale glides above a futuristic coastal city, its body filled with glowing constellations and slow-moving clouds. The camera begins between glass towers, rises quickly, and circles the whale as luminous particles trail from its fins. Golden volumetric light, immense scale, cinematic photorealism, one continuous shot. Audio: deep whale song, rushing wind, distant city ambience, and a restrained orchestral swell.它有效的原因是主体单一、揭示动作明确、镜头连续,而且声音和画面事件相互对应。
A bicycle courier races through a narrow neon market during heavy rain. Handheld camera follows close behind, weaving past steaming food stalls before moving beside the rider for a determined profile close-up. Wet pavement mirrors magenta and cyan signs, realistic spray and fabric movement, tense nighttime atmosphere. Audio: rain on metal awnings, bicycle chain, tires cutting through puddles, distant market voices.A lone explorer approaches a black monolith half-buried in a wind-carved desert. Start with a tight shot of boots crossing rippled sand, then crane upward to reveal the enormous structure against the horizon. Harsh late-afternoon sunlight, long shadows, fine dust moving across frame, grounded science-fiction realism. Audio: dry wind, fabric flutter, low resonant vibration from the monolith.A minimalist glass perfume bottle rests on black volcanic stone. The camera performs a slow 180-degree orbit while a narrow warm light travels across the glass, revealing condensation droplets and the engraved label. Premium commercial composition, dark background, shallow depth of field, crisp product edges, no extra objects. Audio: subtle glass chime, soft stone texture, low ambient pulse.A white performance running shoe floats above a matte red platform. The sole flexes once as the camera pushes in, then the shoe rotates to reveal the tread while a controlled burst of red dust expands behind it. Clean sports advertising style, hard rim light, sharp logo and material detail, energetic but uncluttered. Audio: compressed air release, rubber impact, short bass hit.A premium coffee bag stands on a warm wooden counter beside scattered beans. Morning sunlight moves across the package as the camera slides slowly from left to right. Steam rises behind the bag, curls around the top edge, and clears to leave the label readable. Natural commercial photography, rich brown and amber tones. Audio: quiet café ambience, beans pouring briefly, gentle steam hiss.做商品视频时,优先描述商品周围的镜头、光线和环境运动。除非变形就是创意本身,否则不要要求商品本体扭曲。
Close-up of a tired detective sitting in a parked car at dawn. She watches the empty street, breathes out slowly, then says, “He was never coming back.” The camera makes a subtle push toward her eyes without cutting. Cool blue window light, natural skin texture, restrained performance, realistic reflections. Audio: the spoken line, distant traffic, turn signal clicking softly.A stop-motion inventor in a cluttered workshop tightens the final screw on a tiny brass robot. The robot wakes, looks around, and offers the inventor a glowing flower. Medium two-shot followed by a gentle push-in on their reaction. Handmade clay and brass textures, warm practical lights, playful timing. Audio: tiny gears, a surprised laugh, soft music-box notes.A fashion model in a sculptural silver coat stands in a pale concrete gallery. She turns toward camera as the coat unfolds like metallic petals, then holds a calm direct gaze. Slow dolly-in with a slight clockwise arc, editorial framing, soft overhead light and clean shadows, realistic fabric physics. Audio: fabric movement, spacious room tone, minimal electronic beat.使用对白时,把准确台词放在引号中,并让同时发生的动作保持简单。
Macro shot of a luminous blue butterfly opening its wings on a dew-covered fern at dawn. The camera moves forward slowly while mist drifts through sunbeams and droplets sparkle on the leaves. Natural cinematic color, shallow depth of field, highly detailed wing texture, gentle continuous motion. Audio: quiet forest ambience, delicate wing flutter, distant birds.Inside a dark sea cave, a wave rolls in and briefly fills the rock walls with turquoise reflected light. The camera floats just above the water and moves toward the opening as the wave retreats. Photorealistic water, wet stone texture, high dynamic range, calm documentary style. Audio: deep water surge, droplets falling from the ceiling, distant gulls outside.An old circular library stands intact on the ocean floor. Schools of silver fish pass between the shelves while loose pages rise from an open book and transform into glowing jellyfish. The camera begins above the book and pulls back to reveal the entire room. Dreamlike realism, blue-green caustic light, elegant slow motion. Audio: muffled underwater ambience, paper rustle transformed into soft chimes.Vertical 9:16 close-up of a croissant splitting open to reveal dozens of tiny illuminated rooms inside, each with miniature bakers at work. Start immediately with the crust cracking, push rapidly toward the interior, then hold on the surprising miniature world. Warm bakery light, crisp pastry texture, playful premium ad style. Audio: sharp crust crack, tiny workshop sounds, upbeat one-beat reveal.A red paper airplane flies across a clean white studio, folds itself midair into a red origami fox, lands, and looks directly at camera. The fox leaps upward and unfolds back into the same airplane position for a seamless loop. Locked wide shot, soft shadow, precise paper folds, minimal graphic style. Audio: paper creases, light landing tap, short airy whoosh.上传的图片已经定义人物、构图、服装、商品和色彩。图生视频提示词应该把更多篇幅用于动作、镜头、时间和声音。
Begin exactly from the first image. The camera moves slowly forward as afternoon light fades into blue hour. Leaves and hair respond to a gentle wind while practical lights turn on one by one in the background. Transition continuously toward the composition of the final image, preserving the subject’s identity, clothing, and proportions. End on the last image and hold steady for the final half-second. Audio: light wind, distant city ambience, one soft tonal rise during the transition.不要重新描述参考图中的每个像素。重点说明什么必须保持稳定,以及什么应该发生变化。
| 模式 | 提示词需要控制 | 应避免 |
|---|---|---|
| 文生视频 | 主体、场景、构图、动作、镜头、视觉和声音 | 太多主体或互不相关的动作 |
| 图生视频 | 运动、镜头、时间、声音和必须保持的细节 | 用新场景替换输入图片构图 |
| 首尾帧 | 两张图片之间的连续路径与结尾停留 | 主体或几何关系完全不兼容的两张图 |
准备首尾帧图片前,建议先阅读 H3 Max 分辨率指南。
测试构图时先用 Balanced,方向稳定后再尝试 Quality。
长度应该足以定义一个镜头,但不要包含多个相互竞争的场景。一段涵盖主体、动作、镜头、视觉和音频的短文是很好的起点。
不是。Max H3、H3Max、MiniMax H3 Max 和 H3 Max 是用户常用的不同搜索写法,本页统一使用 H3 Max。
可以在同一提示词中描述对白、环境声、音乐提示和音效。对白应尽量简短,并让关键声音对应画面事件。
480P 适合成本更低的草稿,768P 适合观察细节。完整建议见 H3 Max 480P 与 768P 对比。
两者面向不同工作流。本站的 H3 Max 侧重快速 480P/768P 生成,而 MiniMax H3 是范围更广的模型家族。选择前可阅读 H3 Max 与 MiniMax H3 对比。