Zoom & Object Enlarge · skill pública de Open Skill
Zoom & Object Enlarge
Zoom & Object Enlarge es una Open Skill pública en Digen. Ábrela para ver qué crea este flujo.
Custom
0 0 0
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4150474/2026-10-15/xhzde3jy.jpg
# Zoom & Object Enlarge
Turn an uploaded image into a zoomed-in version in which the main object(s) are enlarged and fill more of the frame. The result must look like the *same* picture, only closer: the colours, materials and lighting stay identical, and the only change is a tighter framing around the enlarged subject.
- **Input**: one reference image (the photo the user uploaded) + optional text saying which object to enlarge and how far to zoom.
- **Output**: one image, produced with image-to-image generation (`model='i2i'`).
- **State**: none. Nothing is stored between turns — keep everything needed in the conversation or on the canvas.
## When to use
- The user uploads a photo and asks to "zoom in", "crop closer", "make the subject bigger", "enlarge the object", "digital zoom", "fill the frame with it".
- The user wants the main object to dominate the composition instead of the wide original shot.
- The user wants to adjust an earlier result (more zoom, less zoom, focus on a different object).
## Steps
1. **Find the reference image.** Use the image the user uploaded in this conversation (its asset id or `s3://` uri). This skill cannot run without a reference image — if the user only wrote text and attached nothing, ask for the photo and stop.
2. **Decide the target subject and the zoom level.**
- *Subject*: if the user names an object ("zoom on the dog"), use it; otherwise pick the single most visually dominant subject in the photo.
- *Zoom level*: derive it from the wording. "Slightly closer" → mild; "zoom in / enlarge" → medium; "very close-up / fill the frame" → strong. Default to medium.
3. **Write the prompt** in English following the conventions below (all four parts).
4. **Call `generate_image`** with `model='i2i'`, passing the source image in `input_urls`.
5. **Reply** in one or two natural sentences (what was enlarged, how far). Do not paste long URLs.
## Prompt conventions
Always write the prompt in English and cover all four parts. The whole image must serve the single theme of "zoom and object enlarge": the enlarged subject, the original colours, the original materials, the original light.
**(a) The zoom action / reframing** — state explicitly that this is a zoomed-in, tighter crop:
- mild: `"slightly closer framing, the main subject a little larger in the frame"`
- medium: `"zoomed-in view, tighter crop, the main subject enlarged and filling most of the frame"`
- strong: `"extreme close-up, the main subject filling the frame, the surroundings mostly cropped out"`
**(b) The subject to enlarge** — name the object and describe it briefly so it stays recognizable:
- e.g. `"the woman's face and upper body"`, `"the red vintage sports car"`, `"the single yellow sunflower"`.
- Keep the subject's identity, shape and details faithful to the reference; do not redraw or restyle it.
- Only the main object is enlarged, so that the composition serves just that one subject.
**(c) Style — must stay identical to the source** — colours, materials and lighting are unified with the original photo; no new look is invented:
- `"same colour palette, same lighting direction and mood, same materials and textures, same photographic/rendering style and grain as the reference image; no new objects, no new colours, no restyling"`.
**(d) Composition / layout** — subject-dominant and clean:
- `"subject-dominant composition, centred on the enlarged subject, tighter crop, background pushed to the edges or partially out of frame, no added borders, no letterboxing"`.
Prompt template assembled from (a)–(d):
> `A zoomed-in, tighter crop of <subject>, enlarged and filling most of the frame. Same colour palette, same lighting direction and mood, same materials and textures, and the same photographic/rendering style as the reference image. Subject-dominant composition, centred on the subject, background partially cropped out. Faithful to the original subject's identity and details; no new objects, no new colours, no restyling.`