AI Video Prompts for Close-Up Video
Macro reveals, extreme close-ups that turn a face into a landscape, product textures that demand to be touched, and the technique of making a single detail carry the weight of an entire scene — these Seedance 2.0 close-up prompts direct the camera at the scale where everything changes.
Close-up video is not about proximity — it is about revelation. The wide shot establishes; the close-up proves. A cityscape says "this is where we are"; a close-up on a specific cracked tile in the sidewalk says "this is what it costs to live here." The scale shift is the meaning, and every genre of video has a version of it. What separates a functional close-up from a great one is not how small the subject is or how close the lens is — it is whether the close-up reveals something that the wider context made invisible, and whether it does so in a frame that has its own internal compositional logic. The structural principle of close-up direction is simple and almost universally ignored in AI video prompts: name what is being revealed, not just what is being shown. "A close-up of a hand" is a framing instruction. "An extreme close-up on the back of a hand, veins raised by cold, the skin at the knuckle creased into a map, a slow pull of focus from the sharpest crease to the soft edge of the frame" is a revelation — the camera is finding something in the hand that the hand does not announce about itself. That difference — the difference between a frame and a discovery — is what close-up direction is for. There are five scales of close-up, each with a distinct spatial grammar and a distinct relationship between subject and frame. The medium close-up (MCU) frames the face from the chest to the crown, keeping the subject in full social context. The MCU is the interview and dialogue register — the camera is as close as a respectful conversation, and the subject occupies the frame's center of gravity without filling it. The close-up (CU) frames from the base of the chin to above the eyebrows, removing the social context and making expression the subject. This is the shot that carries emotional weight in cinema: the single tear, the jaw that tightens, the small involuntary smile. At this scale, the eyes are the compositional axis and every micro-expression is legible. The big close-up or tight close-up fills the frame with the face from chin to forehead, reducing the background to texture and making the face itself a landscape. At this scale, pores are visible, light behaves differently on the micro-geometry of skin, and the subject's relationship to the camera becomes intimate to the point of pressure. The extreme close-up (ECU) isolates a single element — the eye, the mouth, a fingerprint, a reflection in a lens — and makes the scale reversal complete. The ECU tells the viewer: this specific part of this specific thing, right now, is the entire world. The macro close-up moves beyond human subjects into the world of objects and surfaces: the pore of a fabric, the individual crystal in a salt grain, the surface tension of a liquid before it breaks. At macro scale, the camera reveals a world that exists at a scale the eye cannot normally perceive, and the strangeness of the ordinary at that scale is the subject. Lens and depth of field are the two technical choices that determine whether a close-up works. At close-up to ECU distances, depth of field collapses dramatically — an f/1.4 aperture at a face-filling close-up distance produces a focus plane measured in millimeters, which means that sharp focus on the eyes produces soft focus on the ears in the same frame. This selective focus is not a limitation — it is the compositional tool. At the ECU of a face, the eyes should be sharp and the nose tip should fall to soft focus; that hierarchy tells the viewer exactly where to look and confirms that this is a deliberate close-up rather than an awkward crop of a wider shot. At macro distances, a small aperture (f/11 or higher) is often needed just to get a single plane of a small object in focus — but even then, the subject's near and far edges blur to near-abstraction. Naming the focus plane explicitly in the prompt — "sharp focus on the iris, the nose and ear falling to soft bokeh" — gives Seedance the compositional hierarchy to execute rather than leaving focus distribution to default. The lighting grammar of close-up video differs from every other scale because the subjects are small enough that the light source's direction relative to the surface determines the entire topography. At MCU distance, standard three-point lighting works normally. At CU and ECU distance, a single off-axis key light creates shadow structures across the face that reveal the micro-geometry of the skin: the hollow under the cheekbone, the slight asymmetry of the orbital ridge, the way light catches the upper lip differently from the lower. Naming the key light direction and its softness at close-up distance gives the model the tools to produce a face as a topographical record rather than a lit surface. For product and material close-ups, the light direction determines whether the surface reads as texture or as flat color: a raking light at 20 degrees to a rough surface grazes across every imperfection and makes the texture three-dimensional; a front fill at 0 degrees fills all the shadows and makes the same surface read as flat. The difference is a camera-to-light-angle specification, and making that specification explicit is what separates product close-up video that makes the material irresistible from video that merely shows it. Motion at close-up scale operates differently from motion at wide scale because even small camera movements produce large compositional shifts. A 2cm rack focus across a face at ECU distance completely changes the compositional axis. A 1mm dolly at macro distance changes the entire depth structure of the frame. This means that close-up motion has to be named at a finer resolution than wide-shot motion: not "a slow push-in" but "a 3cm push-in over 4 seconds, maintaining sharp focus on the left eye as the right eye moves slightly past the critical focus plane." The motion that works at close-up scale is the motion of discovery: a slow rack of focus revealing successive layers of a subject, a micro-pan that finds the specific texture detail at the edge of frame and centers it, a pull-back that starts inside an unidentifiable texture and widens until context reveals what it was. These are not just camera moves — they are the structural arc of what the viewer is shown: ignorance of context → recognition → understanding. Across all five scales and all subject types, close-up video has one governing principle: the frame has to earn its proximity. A close-up that is merely a cropped wide shot earns nothing. A close-up that reveals something the wider view could not — the tension in a jaw that the dialogue does not express, the surface texture of a material that explains its cost, the specific light reflection in an eye at the moment of a decision — that close-up carries meaning that no other framing could have produced. Every prompt in this gallery is built on that principle: the subject, the scale, the focus plane, and what the close-up is specifically revealing. Copy a prompt, name what your close-up is for, and keep the focus hierarchy and light direction intact.
More use cases
Frequently asked questions
What are the best AI video prompts for close-up video?
The best close-up prompts name three things together: the specific element being isolated (the iris, the fabric pore, the product surface texture, the micro-expression), the focus plane and depth of field instruction (sharp focus on the left eye, the surrounding face falling to soft bokeh), and the light direction relative to the surface (raking side light to reveal texture, soft front fill for skin, hard backlight for product translucency). Without all three, the model defaults to a generic face or object at close range with no compositional logic. Every prompt in this gallery uses that three-layer structure.
How do I write an extreme close-up (ECU) prompt for Seedance 2.0?
Structure the ECU prompt in four elements: (1) the specific subject element — not "a face" but "the left eye, upper eyelid slightly lowered"; (2) the focus plane instruction — "sharp focus on the iris, the lashes at the near edge just slightly soft, the bridge of the nose falling to bokeh"; (3) the light direction and quality — "soft box from camera-left raking slightly across the eye socket to reveal the micro-geometry of the orbital area"; (4) the motion if any — "a 2cm pull-back over 4 seconds, maintaining critical focus on the iris as the surrounding face slowly enters frame at the edges." The focus plane specification is the most commonly missing element and the one that most determines whether an ECU looks deliberate or accidental.
Can Seedance 2.0 generate macro and product texture close-up video?
Yes. Seedance 2.0 handles macro and product close-up video well when the prompt names the specific surface element, the light direction that reveals texture, and the focus plane. For product texture, raking light at a low angle to the surface — "a raking side light at 15 degrees to the material surface, grazing across the weave and creating a shadow structure that makes each thread three-dimensional" — is the most reliable technique for making surfaces feel touchable on screen. For macro work, naming the specific visual phenomenon you want revealed (the surface tension before a liquid breaks, the crystal structure of a material, the micro-scale pore pattern of a fabric) gives Seedance a specific discovery to execute rather than a generic small-scale shot. Browse the close-up prompts here for real examples with preview videos.