CommunityImagegithub.com

photo-to-paper-collage

Turn any uploaded photo (landscape, street, architecture, portrait, pet, food, interior) into a collage poster that blends "half real photo, half hand-drawn on paper." Light, water, sky, outlines, and textures from the original image cross the paper's edge and become a plausible action or object on warm ivory paper, or the photo appears like a torn window embedded in the paper. The user can simply say "make one with photo-to-paper" and the system will ideate and output an image. Suitable for photo-to-illustration, semi-real semi-hand-drawn, photo-into-paper, collage poster, and real-scene extended hand-drawing.

What is photo-to-paper-collage?

This skill transforms any user-uploaded photo into a unique collage poster that merges half of the original photographic reality with a hand-drawn extension on warm ivory paper. The core creative rule is that one continuous element from the photo—such as a body of water, sky, light beam, or texture—crosses the paper's edge and transforms into a coherent object or action on the drawn side, maintaining its original color and texture. The AI agent must generate a fresh concept for each image, selecting from three approaches: a miniature figure interacting with the borrowed element (e.g., painting with a lake's sunset), the borrowed element becoming an object (e.g., ocean light turning into a sail), or the photo itself becoming a window or torn page. The skill provides detailed guidance on composition, hand-drawing style (pencil or ink with minimal color), and a structured prompt template for tools like Doubao Image Generation, GPT Image, or Seedream. The output checklist ensures quality: the original photo remains unaltered, the transition point is clear, the transformed element is visually substantial, and the paper area mostly stays clean with only one hand-drawn change and a small caption.

Works with~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/lindaifeng/photo-to-paper-collage/tree/HEAD

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

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

Photo-to-Paper Collage

The user provides any photo, and you conceive and deliver a finished image: part of it is the real scene from the original photo, and the rest is hand-drawn on warm ivory paper, with something happening between them that seems "impossible at first glance, but makes sense upon closer look."

Each photo requires its own unique concept—do not reuse fixed compositions. The examples below only illustrate possibilities; the final content, composition, and variation should grow from this specific photo.

When the user uploads only a photo with no idea, follow the four steps below to decide on everything yourself and output the image directly—no need to ask questions.

What This Effect Is: Five Unchanging Features

Regardless of the photo, the finished image must satisfy all five points:

  1. Real photo + real paper surface: About half is the original photo, kept real—no redrawing, no color changes; the other half is warm ivory art paper with subtle paper grain.
  2. Borrow something across the paper edge: Something that already exists in the photo (a beam of light, water, sky, an outline, a texture) crosses the photo's edge into the paper area.
  3. The material stays continuous: Once on the paper, it changes shape but not material. A sunset reflection in a lake flows down as paint, but the paint is still sunset-colored; a golden sea surface becomes a sail, but the sail still carries the same shimmering light.
  4. It becomes something plausible: On paper, it turns into an object or is used in an action by a small figure. This transformation relates to the original photo—similar shape, logical use, or matching mood.
  5. Quiet and restrained: Hand-drawing uses pencil or ink lines; colors are borrowed only from the original photo; there is only one transformation, the rest of the paper is blank; and a small caption is included.

Step 1: Read the Image, Find What to "Borrow"

First, describe the photo: what is the subject, where does the light come from, what are the most beautiful colors, what continuous surfaces or lines exist. Then pick one thing best suited to cross the paper edge:

Common in PhotosWhat Can Be Borrowed
Lakes, seas, rivers, puddlesReflections, shimmering light, wave lines, shorelines
Sky, sunset, cloudsThe sky's own color and cloud patterns, a ray of sunset light
Mountains, rocks, terraced fields, dunesRock textures, ridge lines, layered steps
Buildings, lighthouses, towers, bridgesSpires, towers, eaves lines, arches, window frames
Streets, train tracks, paths, railingsExtending lines, perspective directions
Trees, flowers, vines, grassBranches, petals, fallen leaves, shadows
Lights, night scenes, neonLight patches, light strips, reflections of lights
Portraits (people in the photo)Clothes hem, scarf, hair, shadows, objects in hand
PetsTail, fur direction, shadows, leash
Food, coffee, interiorsSteam, vapor, rim light on cups, window light, tablecloth textures

Selection criteria: Best color or texture, continuous shape, recognizable at a glance on a small phone screen.

Borrow surfaces, not thin lines or faint light. Prioritize borrowing a large area from the photo—one with rich color, full saturation, and clear texture: an entire lake surface, the whole sky, a full snowy path, a whole curtain, a whole expanse of sea light. A faint ray of light or a thin line, when crossed to paper, will become an indistinct thin strip that cannot carry the image. If the photo has only a thin line to borrow, let it expand into a surface on paper (a train track unfurls into a film roll; a light beam spreads into a sail).

Step 2: Choose a Connection Method, Imagine What It Becomes

Choose one of three methods. Which one to pick depends on the photo: if the photo has a continuous surface (water, sky, fabric) → use A or B; if there is a clear anchor point (spire, mast, branch, cup rim) → use B; if there are too many details making transformation difficult, or if a person is the main subject → use C.

A: A Figure Takes Action

Draw a small figure on paper performing an action with the borrowed thing, causing it to cross the paper edge.

  • The action must match the material: liquids and light can be painted (with a brush tip drawing them down), scooped, fished; sheet-like water or sky can be lifted, pulled, folded, hung to dry; lines can be pulled, sewn, plucked; light things can be blown, set free.
  • The contact point must be physically connected: the brush tip rests at the end of the paint, hands grip the gathered folds of fabric, the fishing line touches the water.
  • The figure is small (about 15–30% of the image height), shown from behind or in profile, with no clear facial features; pencil or ink lines, possibly with light colors; body posture matches the direction of force (leaning back when pulling, reaching up when painting).
  • Example: Sunset lake → the reflection narrows downward, flowing into a stream of paint, ending at the brush tip of a cross-legged figure painting. Blue lake → the whole reflection is lifted like fabric, its folds gathered into the hands of a girl wearing a straw hat. Coffee photo → the steam from the cup's rim is caught on paper by a small figure using a net.

B: Scene Becomes an Object

The borrowed thing transforms into an object on paper, which hangs from, supports, or rests on a real anchor point from the original photo.

  • The object must "owe" something to the original photo: the same texture (sea light → sail cloth, sky → umbrella canopy, rock layers → inner wall of a stone vessel), the same shape (spire → center of a parasol's ribs, arched bridge → swing, train tracks → an unfurled film roll), or a logical function (holding, shading, carrying, illuminating).
  • The object's frame (ropes, parasol ribs, wooden poles, stone rims) is drawn with fine pencil or ink; the surface is covered with the original photo's material.
  • The anchor point must be identifiable: the rope is tied to a real mast, the parasol ribs grow from a real spire, the stone vessel extends from a real rock layer.
  • A very small figure can be added for scale (sitting on the vessel's rim, standing under the parasol), but should not dominate the scene.
  • Example: Sunset sea with sailboat → a rope descends from the real mast, suspending a sail made of golden sea light that falls onto paper. Lighthouse → from the top radiate 8-10 parasol ribs, the canopy being the original photo's blue sky. Mountains → the nearest rocks form a stone vessel holding the brightest golden part of the sky. Street train tracks → the tracks extend onto paper, becoming a hand-drawn pencil line that recedes into the distance.

C: Real Scene as a Window

Don't transform anything—just let the photo itself become part of the paper: a torn paper hole, a turned page, a fold, a drawn window.

  • The window's shape follows the scene: the torn edge follows the skyline or mountain ridge, the turned page follows the shoreline.
  • Add a little foreground on the paper side: silhouettes of leaves reaching in from outside the photo, the binding crease of a sketchbook, pencil lines of a window frame.
  • Use this method when a person is the main subject: the person stays whole in the window, their identity and appearance unchanged.
  • Example: City lake view → a hand-torn paper hole in the middle reveals the skyline and boats; dark leaf silhouettes reach in from the top-left and top-right corners. Portrait → the person remains inside a pencil-drawn arched window; a few strokes of street scene continue outside the window on paper.

Regardless of Method: The Transformation Must Have Substance

  • The transformed thing is the image's main subject: It must occupy at least one-third of the paper area, spreading wide, hanging down, rolling up, or opening wide—not a small decoration in a corner.
  • It should be as clear and vibrant as the photo: Its surface carries the original photo's color, texture, and light (the sail has distinct shimmering waves, the fabric has clear reflections); clarity is close to the photo, with pencil/ink lines only at the edges and framework. The more blank the paper, the more vibrant the transformation should be.
  • No new content drawn on its surface: The transformed object's surface only carries the original photo's texture or image—no additional patterns, landscapes, or drawings on top.
  • Figures and frames are supporting players: Figures are small, frame lines are thin; they are there to "explain how it connects," while the visual weight is on the transformed object.

Step 3: Determine the Composition

  • Aspect ratio: Follow the original photo's orientation. Portrait → 3:4, landscape → 4:3 or 5:4, near square → 1:1. Use the user's specification if provided.
  • Photo placement: For photos with sky and distance at the top, place the photo on top and the paper below; for vertical landmarks, place the photo on one side and the object extends to the other; for method C, the photo is in the middle.
  • Proportion: The photo occupies about 45–55%, and most of the remaining paper is left blank.
  • Transition edge: Can be a clean horizontal line, a line following the scene's contour, a hand-torn rough edge, or a soft fade. Choose the one that best matches the image, and use only one type throughout.
  • Caption: One line, 2–6 words, placed in the blank area, small and light. Choose one of two styles: a short English hand-written phrase (e.g., Paint the sky, A sail made of sunset.), or a short Chinese poem in Song typeface (e.g., 借群山,盛一盏晚霞。). The caption describes the event in the image. Omit if the user says no text. No watermarks or signatures.

Step 4: Write the Generation Prompt and Output the Image

Compose the concept into a complete generation prompt, and pass it together with the original photo to an image generation or editing tool (e.g., Doubao Image Generation, GPT Image, Seedream). Each item should refer to the specific content of this photo:

Using this photo as the base, create a {aspect ratio} "photo-to-paper" collage poster.
Keep: {photo placement, percentage}, the original photo's {specific scene, light, and color} remain real—no redrawing, no color changes.
Transformation: The {borrowed area} from the original photo crosses the paper edge, {how it gradually transforms}, becoming {an object / an action performed by a figure} on paper; it is the main subject, occupying over one-third of the paper area, its surface carries the original photo's {texture and color}, as clear and vibrant as the photo, with no new content drawn on it.
Connection point: {specific contact, e.g., "the brush tip touches the end of the paint", "the rope is tied to the real mast"}, must be visible on a small phone screen.
Paper: warm ivory art paper, fine paper grain; {description of the figure/object, pose, position}; pencil or ink line art, only a hint of the original photo's colors, the rest left blank.
Caption: {position} a small, light line of {hand-written English / Song typeface Chinese} "{caption text}".
Do not: redraw the photo area, change the sky color, add heavy shadows, sticker-like white borders, multiple unrelated transformations, extra text, or watermarks.

Example: User uploads only a photo of coffee by the window

Using this photo as the base, create a 3:4 portrait "photo-to-paper" collage poster.
Keep: The top 50% of the image is the original photo—the window light, wooden table, and coffee cup remain real, no redrawing, no color changes.
Transformation: The entire pleated cream curtain from the photo crosses the bottom edge, along with the window light's shading in the folds, falling onto paper; the folds become looser downward, transforming into a long skirt that spreads across half the paper; the skirt carries the curtain's fabric texture, folds, and warm golden light, as clear as the photo.
Connection point: The skirt's waist is exactly the real curtain at the photo's bottom edge—no break.
Paper: warm ivory art paper; at the skirt's hem, a very small girl in profile lifts the hem with both hands and walks forward; pencil lines, the figure gets only a touch of light brown; the rest is blank.
Caption: Lower left, hand-written English: "Wearing the morning light", small and light.
Do not: redraw the photo area, change the sky color, add heavy shadows, sticker-like white borders, multiple unrelated transformations, extra text, or watermarks.

When you have an image tool: Write the prompt, generate the image, deliver the final image, and add one sentence describing what was borrowed and what it became. When you do not have an image tool: Deliver this prompt and tell the user to send it together with the original photo to a tool that supports "reference image generation."

Post-Generation Self-Check (If any check fails, regenerate only for that issue)

CheckWhat Failure Looks LikeHow to Fix the Prompt
Can the original scene be recognized?Photo area has been redrawn as an illustration, colors changedEmphasize "keep the original photo, no redrawing, no color changes," specify the location and proportion of the photo area
Is the material continuous?The object on paper has changed color, doesn't look like it came from the photoWrite "{object}'s material is the original photo's {texture}"
Is the connection point clear?Hand doesn't touch, rope is floating, broken into two piecesDescribe the connection point as a concrete contact action
Does the transformation have substance?The transformed object is a faint thin strip, a small cornerBorrow a full surface instead; write "occupies over one-third of the paper, color and texture as clear and vibrant as the photo"
Is the surface from the original photo?New patterns or small drawings added on the transformed objectWrite "surface only carries the original photo's {texture}, no new content drawn on it"
Is there only one transformation?Both a figure and a torn edge and an objectDelete extraneous transformations, keep only one
Does it make sense?The object on paper has no relation to the photo, could be placed in any photoReplace with an object that relates to the original photo's shape, texture, or function
Is there enough blank space?The paper area is filled up, looks messyWrite "the rest of the paper is blank, with only {one} hand-drawn element"

Related Skills