Community图像github.com

Mumu888666/photoreal-human-portraits

把人物设定或参考图转为有真人摄影质感的 AI 人像。适用于写实角色定妆、头肩照、面部特写,以及修正磨皮、假眼神、假发感或光影失真的现有角色图。可交付完整提示词,或在用户要求出图时生成并按成图验收。

photoreal-human-portraits 是什么?

photoreal-human-portraits is a Claude Code agent skill that 把人物设定或参考图转为有真人摄影质感的 AI 人像。适用于写实角色定妆、头肩照、面部特写,以及修正磨皮、假眼神、假发感或光影失真的现有角色图。可交付完整提示词,或在用户要求出图时生成并按成图验收。.

兼容平台~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/Mumu888666/photoreal-human-portraits/tree/HEAD

Installed? Explore more 图像 skills: steipete/songsee, affaan-m/frontend-design, mattpocock/improve-codebase-architecture · View all 6 →

在你喜欢的 AI 中提问

打开一个已预加载此 Agent Skill 的新对话。

文档

Photorealistic Human Portraits

Imagine the person has just sat down in front of the camera, looking at a specific target. First, determine the facial structure, gaze, and lighting, then distribute skin details across the face. The character's description determines their appearance and state; the common qualities of reference images determine the photographic texture.

From Requirements to the Image

  1. Lock in the character and the task. Extract age range, face shape and notable features, hairstyle, clothing, the character's current situation, frame composition, and purpose. Distinguish whether a reference image is for the same character, borrowing photographic texture, or referencing only one part. When borrowing only texture, write the new character's appearance separately to avoid bringing the reference's face, scars, heterochromatic eyes, or clothing. For the same character changing scenes or expressions, keep facial proportions, eye shape, nose/lip structure, hairline, and identifiable marks. When reversible photographic parameters are missing, choose reasonable defaults and briefly explain the choice when delivering.
  2. Use actions that could be photographed to describe emotion. Write what the character is looking at or waiting for in this moment, then choose one or two visible changes, e.g., "Looking at the interviewer beside the camera, waiting for them to finish, jaw relaxed, upper eyelids slightly raised, lips lightly together." Also specify the shared target of both eyes, head angle, and mouth state. Alertness, fatigue, and calmness each have different eyelid and facial states; choose according to the character profile.
  3. Establish a readable photographic base first. When no specific purpose is given, default to eye-level single-person head-and-shoulders framing, a clean gray or gray-blue background, soft main light slightly off-center, with moderate shadows retained in the eye sockets and cheekbones. Leave space for the hairstyle and shoulders in the frame; focus on the eyes while keeping the nose, mouth, and cheeks readable. When skin inspection or character consistency is needed, use a close-up of the same character or crop the original for verification. Scene shots, full-body shots, or narrative frames are composed per user requests, keeping visible details appropriate to the final image scale.
  4. Write the prompt as a ready-to-use complete set of instructions. Order: character & scene, current state & gaze, zoned skin & texture, lighting & camera, necessary exclusions. When no model is specified, use natural language; do not guess model parameters. Extract exclusions only if the tool has a dedicated negative prompt field. When only the prompt is requested, deliver the complete prompt directly. When image generation is requested, use available image tools to generate and review the image. When the user wants to fix an existing image, use the original as an editing reference, specifying only the areas to change and the character traits to keep. If the tool cannot maintain identity using the reference image, distinguish between new generations and identity‑consistent variants.

Place Details Correctly

Body PartWhat the Prompt Should Control
Skin undertoneDescribe gentle color variations. The nose wings, eye area, and cheeks may be slightly pink, warm, or darker depending on the character's state. Keep natural skin tone in shadows; avoid a uniform orange-pink coating.
Pores and fine linesOnly write visible areas. Close-ups may show subtle texture on the nose wings, nose sides, and forehead, with shallow creases along the eyelid direction under the eyes. For mature characters, add age lines between the brows or under the eyes; for younger characters, keep details light, localized, and sparse. Optionally add a few bits of vellus hair, light marks, scattered small spots, or stubble; avoid uniform face-wide speckling.
ShineThe nose tip and lip surface may have short, soft bright highlights; cheek highlights are broad and weak; keep shadows in the eye sockets. Shine intensity is governed by the same lighting setup and also respects the character's makeup and skin condition.
EyesBoth eyes aim at the same target. The upper eyelid naturally covers part of the iris according to the viewing angle; the iris, pupil, sclera, and eyelid margin are clearly layered. Sclera retains some gray; the eyelid waterline is slightly moist; the reflection in the eyes matches the main light position. The opening of the two eyes may differ slightly, but the gaze direction remains coordinated.
Mouth and facial tensionUse lip line openness, corner tightness, and jaw tension to express state. The lip surface has fine lip lines; color and gloss vary slightly across areas. Avoid outlining the mouth with hard edges and heavy bright spots.
Hair and surrounding materialsThe hairstyle first has grouped volume, then adds root color variation and a few stray hairs. Eyebrows and stubble grow from their respective areas. Fabric has fiber and creases; metal accessories have local reflections; makeup edges adhere to the skin and eyelids.

Arrange details according to the frame scale. Head-and-shoulder shots prioritize facial structure, gaze, skin tone layering, and hairstyle silhouette. Facial close-ups may require readable waterlines, fine lines, vellus hair, or pores. Avoid using global sharpening, film grain, or high contrast as substitutes for real texture. Retain an appropriate amount of natural asymmetry; avoid mirrored faces, identically round and wide eyes, template smiles, or unfocused eyes.

Image Acceptance and Repair

Look at the whole image first, then zoom in on the eyes, nose, cheeks, lips, and hairline. Locate issues in the order below; do not fixate only on sharpness.

  • Character. Age, bone structure, hairstyle, makeup, and character temperament match the requirements. Variants of the same character should be compared with the reference image for eye/brow proportions, nose/lip structure, hairline, and identifiable marks.
  • Eyes. The gaze target and state should be recognizable even from a thumbnail. When zoomed in, check gaze direction of both eyes, eyelid coverage, iris-to-sclera transition, and whether reflections match the light source. If the eyes look glassy, first reduce sclera brightness and highlights, and correct the shared point of focus.
  • Skin. The face retains structure and skin tone layering even in a thumbnail. Zoomed in, visible zoned texture matches the frame scale, age, and makeup. If the skin looks waxy, first restore localized matte areas and slight color variation. If it looks sandpapery, first reduce uniform, full-face texture, then put details back into a few areas only.
  • Lighting and materials. Eye reflections, nose tip highlights, nose shadow, and facial shadows should be explainable by a single lighting setup. Hair, fabric, and accessories each have their own reflections and details. Remove playback buttons, cursor pointers, and other interface overlays that are part of a reference screenshot.
  • Prioritize fixing the most artificial part; after fixing, recheck character identity, eyes, and lighting. If the main issues remain obvious, honestly mark it as needing repair. When only a prompt was written, deliver the prompt and mark the final image effect as pending verification.

Ready-to-Use Example

Request: "42-year-old female detective, after an overnight shift doing a short debrief, tired but alert, realistic ID photo."

A single-subject realistic ID photo, eye-level head-and-shoulders framing, gray-blue background. 42-year-old female detective, short black hair with a few stray hairs at the hairline, dark gray casual jacket, slight fatigue under the eyes. She has just heard a colleague to the right of the camera ask a key question; both eyes are looking toward the nearby colleague on the right, her head turned only slightly, jaw relaxed, upper eyelids slightly raised, lips naturally and lightly together, expression restrained. Keep subtle skin tone variation between the cheeks and nose wings, shallow lines under the eyes following the eyelid direction, low-contrast local texture beside the nose, and fine lip lines on the lip surface. The nose tip has a soft bright highlight; cheeks are mostly matte. The iris is partially covered by the upper eyelid naturally; the sclera is slightly gray; each cornea retains a small reflection consistent with the main light direction. Soft light enters from the upper left of the frame, with slight fill light on the other side; the eye sockets and cheekbones still have soft shadows. Eyes are sharp, nose and mouth readable, background slightly blurred, the jacket retains natural fabric weave and creases. Avoid skin smoothing, full-face grain, plastic highlights, overly bright sclera, mirrored facial features, and on-screen text.

Visual Basis of Original Sample Images

  • Figures 1–2 are a close-up and a head-and-shoulder shot of the same mature male. In the close-up, the glabellar lines, shallow lines under the eyes, stubble, and lip lines vary by area. The eyes look almost directly at the camera; the brows are slightly gathered; the eye socket shadows and lightly closed mouth together form a rather serious state. The wider frame relies on bone structure, hairstyle silhouette, and the suit material to work.
  • Figures 3–4 are a head-and-shoulder shot and a close-up of the same blue-haired female. The close-up shows slightly reddened nose and cheeks, a few scattered small spots, lower eyelid margin, and fine lip lines. The purple-red iris and obvious eyeliner belong to the character's styling; the iris is partially covered by the upper eyelid and retains small reflections; the face still has skin layering. Dark roots and stray hairs crossing over the eye edge complete the hair's realism.
  • Figure 5's blonde female has generally smoother skin but still shows color variation on the cheeks, nose wings, and lip area. The fringe has grouped hair and a few loose strands; both eyes clearly face the camera; the mouth tension is light. Realism can come from subtle color variation and gaze, without adding coarse texture.
  • Figure 6's dark-haired male relies on a localized red mark on the forehead, nose wing reflections, sparse stubble, eyelid opening/closing, and slightly parted lips to distinguish material and expression. The shirt's creases and bead strings also retain their own shape and reflections.
  • The above records visible features of the finished images. The models used by the original author, the original prompts, lighting equipment, and post-processing steps cannot be confirmed from these screenshots.

相关技能