This is the sixteenth original article on aogl.cn, sourced from original/sports-lens/. The folder codename sports-lens refers to a telephoto / viewport framing exercise: use a cave mouth as a natural vignette, lock a terraced cherry-blossom valley with river and mist bands inside that irregular frame, then export MP4s to see whether the frame-in-frame story still holds in motion. It is not provincial tourism marketing, carries no verifiable GPS, and does not teach one-click prompts for any vendor app—only filenames, composition choices, and how this line differs from other “window” archives on the site.

What is on disk

• Gemini_Generated_Image_jib1odjib1odjib1.png — hero still: cave foreground, terrace valley, distant peaks, dawn/dusk sky (cover art for this URL)
• Gemini_Generated_Image_1frod41frod41fro.png — exploration A: heavier rock lip, different fog density
• Gemini_Generated_Image_1frod41frod41fro1.png — exploration B: cherry saturation and water specular tweaks
• video.mp4 — main composite (re-exported 2026-06-22)
• video-2.mp4 — second pass comparing shadow crush and cloud drift

Headings and captions use honest phrases: cave frame landscape, terrace rice fields, cherry blossom valley, natural window composition, frame-in-frame video—each tied to pixels on disk, not unrelated SEO stuffing.

Why a cave frame instead of full-bleed landscape

The site already hosts a high-speed train window line, an airliner cloudscape window, and an apartment 360° pack. This article adds an organic matte: the opening is not rectangular glass but jagged stone, foreground is near-black, and the viewer reads “a limited view” before absorbing the valley. Psychologically, peeking at a hidden Shangri-La beats a flat wallpaper because concealment precedes reveal.

I kept a small cluster of pink-white wildflowers on the cave floor—not botanical survey data, but a scale anchor in absolute shadow. Without it, the vignette feels like a filter; with it, the brain accepts camera-in-cave / subject-outside depth.

Hero still jib1od: why the cover uses this file

Among three PNGs, jib1od has the strongest edge contrast: hanging vines and rock teeth carve the sky into an irregular arch; the valley mid-ground is brightest; distant peaks plus low cloud form a second horizon. The 1frod pair stays local when cherry shifts magenta or fog gets soup-thick—they are regression tests, not carousel heroes. Thumbnails at 16:9 must scream “frame + valley,” not “pink forest wallpaper.”

Two 1frod explorations: what I am comparing

Generative frames often fail when the cave lip drifts—rock thickness changes and the whole perspective lie collapses. 1frod and 1frod1 are two keeps from the same art direction:

• 1frod — wider rock border, more canvas in shadow; tests whether vignette eats the subject
• 1frod1 — brighter river specular, denser blossom; tests whether highlights look like fake HDR

Neither replaces jib1od as hero because articles need one default narrative frame; explorations serve maintainer before/after only.

Valley vocabulary: terraces, blossom, river, mist

Mid-frame terraced fields with water mirrors provide rhythmic greens; I deliberately avoid naming real counties because files are not geotagged. Cherry blocks add a seasonal color chord against green steps so the interior of the frame has a second visual beat under the arch.

Distant peaks and low cloud supply depth. Terrace alone would flatten the cave matte into a sticker; a third atmospheric layer sells walkable space instead of illustration.

video.mp4 and video-2.mp4

Stills lie easily; video exposes lip flicker, blossom boiling, or looping fog tiles. video.mp4 is the 2026-06-22 re-export; video-2.mp4 keeps an earlier parameter set to compare shadow grey versus clipped highlights on the river.

How this splits from other “window lines” on aogl.cn

• Train window — manufactured rectangle + motion-blur exterior
• Sky-plane window — oval porthole + high-altitude cloud deck
• sports-lens — organic rock frame + ground-level valley (this page)
• Travel-through phone — screen rectangle + layered PNG parallax

All four share acceptance tests: is the frame understood before the scene? does the frame stay believable when scenery swaps? Failing takes never enter git.

Tools and scope boundary

Stills came from Gemini image exploration; filenames keep the Gemini_Generated_Image_… prefix like girl1 contact sheets and role-girl cards for batch traceability. I do not reproduce vendor prompts here or sell “generate Shangri-La in one click.” Documentary forks should swap in verifiable locations and EXIF, not this fictional valley.

Post-generation work is local crop and video export only; the HTML article states scope honestly—aligned with AdSense strategy where articles/ is primary content and each URL maps to an inspectable original/ folder.

Maintenance contract

1. Change hero PNG → rerun npm run generate-article-covers and bump dateModified.
2. New explorations → gallery only unless the frame story improves clearly.
3. Do not republish the same figures in hub-links or briefs; long-form lives in articles/ only.

A seventeenth article in the “frame line” would need a new artifact—maybe a ruined tower window or bridge arch—plus MP4 proof before another scope paragraph. I will not clone this template just to inflate article count.