> ## Documentation Index
> Fetch the complete documentation index at: /llms.txt
> Use this file to discover all available pages before exploring further.

# Gallery Tunnel

> An infinite corridor of panels: a square tunnel receding to a vanishing point, its floor, ceiling and walls tiled with saturated colour slabs and caller-supplied images that swell toward camera and exit past you, the far end dissolving into fog.

# Gallery Tunnel

An infinite corridor of panels: a square tunnel receding to a vanishing point, its floor, ceiling and walls tiled with saturated colour slabs and caller-supplied images that swell toward camera and exit past you, the far end dissolving into fog.

**Gallery Tunnel**

```html gallery-tunnel.html
<!doctype html>
<html lang="en">
  <head>
    <meta charset="utf-8" />
    <meta name="viewport" content="width=1920, height=1080" />
    <title>Gallery Tunnel</title>
    <script src="https://cdn.jsdelivr.net/npm/gsap@3.14.2/dist/gsap.min.js"></script>
    <script src="https://cdn.jsdelivr.net/npm/three@0.147.0/build/three.min.js"></script>
    <style>
      *,
      *::before,
      *::after {
        margin: 0;
        padding: 0;
        box-sizing: border-box;
      }
      body {
        background: #000;
        overflow: hidden;
      }
      #gt-root {
        position: relative;
        width: 1920px;
        height: 1080px;
        overflow: hidden;
      }
      #gt-backdrop {
        position: absolute;
        inset: 0;
        background: #000000;
      }
      #gt-canvas {
        position: absolute;
        top: 0;
        left: 0;
        width: 1920px;
        height: 1080px;
      }
      /* Off-screen decode well: the <img> elements live in the DOM so the page
         load event waits for them, but they are never composited themselves —
         WebGL reads them as textures. The well is clipped to 1px; the images
         inside keep their natural layout size on purpose, because texImage2D
         uploads an <img> at its LAID-OUT size — sizing them down with CSS
         uploads a 1x1 texture and every panel samples one dark pixel. */
      #gt-assets {
        position: absolute;
        top: 0;
        left: 0;
        width: 1px;
        height: 1px;
        overflow: hidden;
        opacity: 0;
        pointer-events: none;
      }
    </style>
  </head>
  <body>
    <!--
      Content slots. Every panel in the corridor is a slot, and `images` fills a
      share of them with your own artwork — drop the files in the project's
      assets/ and list them, either by editing the `images` default below or from
      the host composition:

        <div
          data-composition-id="gallery-tunnel"
          data-composition-src="compositions/gallery-tunnel.html"
          data-start="0" data-duration="12" data-track-index="0"
          data-width="1920" data-height="1080"
          style="--images: assets/shot-1.png, assets/shot-2.png, assets/shot-3.png; --imagemix: 0.7"
        ></div>

      Any number of images works; they are dealt to slots by hash, so one image
      recurs at different depths rather than appearing once. Each is contain-fit
      inside its slot, so a 16:9 screenshot keeps its aspect. `imageMix` sets the
      image-vs-colour split (1 = every filled slot is an image), `fill` how many
      slots are filled at all.
    -->
    <div
      id="gt-root"
      data-composition-id="gallery-tunnel"
      data-root="true"
      data-width="1920"
      data-height="1080"
      data-start="0"
      data-duration="12"
      data-composition-variables='[
        {"id":"images","type":"string","label":"Image slots: comma-separated paths, tiled into the panels (e.g. assets/shot-1.png,assets/shot-2.png)","default":"","placeholder":"assets/shot-1.png,assets/shot-2.png"},
        {"id":"imageMix","type":"number","label":"Image share of filled slots (0 = all colour, 1 = all images)","default":0.5,"min":0,"max":1,"step":0.05},
        {"id":"speed","type":"number","label":"Flight speed","default":3,"min":0,"max":8,"step":0.1,"unit":"segments/s"},
        {"id":"palette","type":"string","label":"Slab palette: comma-separated hex colours","default":"#FF6A00,#AB54F7,#EA3737,#0072E3,#00AA3C,#FFB200"},
        {"id":"grid","type":"number","label":"Grid density: slots per wall face per axis (2 = 4 per face, 16 per segment)","default":2,"min":1,"max":4,"step":1},
        {"id":"fill","type":"number","label":"Slot fill rate","default":0.5,"min":0,"max":1,"step":0.05},
        {"id":"fog","type":"number","label":"Fog distance: depth at which the corridor reaches pure backdrop","default":14.25,"min":3,"max":14.5,"step":0.25,"unit":"segments"},
        {"id":"edgeColor","type":"color","label":"Edge line colour","default":"#B0B0B0"},
        {"id":"edgeOpacity","type":"number","label":"Edge line opacity","default":0.5,"min":0,"max":1,"step":0.05},
        {"id":"backdrop","type":"color","label":"Backdrop and fog colour","default":"#000000"}
      ]'
    >
      <div id="gt-backdrop"></div>
      <canvas id="gt-canvas" width="1920" height="1080"></canvas>
      <div id="gt-assets" aria-hidden="true" data-layout-allow-overflow></div>

      <!-- Driver clip: gives HyperFrames a timed element to own on track 0. -->
      <div
        id="gt-drv"
        class="clip"
        data-start="0"
        data-duration="12"
        data-track-index="0"
        style="position: absolute; width: 1px; height: 1px; opacity: 0; pointer-events: none"
      ></div>
    </div>

    <script>
      (function () {
        var DUR = 12;
        var W = 1920;
        var H = 1080;

        // Fixed tunnel geometry, measured off the reference.
        var TUNNEL_WIDTH = 2;
        var TUNNEL_HEIGHT = 1.8;
        var SEGMENT_DEPTH = 1;
        var NUM_SEGMENTS = 15;
        var LINE_RADIUS = 0.003;
        var FOV = 75;
        var INSET = 0.94; // slab gutter inside its cell, so tiles read as discrete
        var SKIN = 0.002; // push slabs just behind the wall plane so edges stay on top

        var ROOT = document.getElementById("gt-root");
        var CS = getComputedStyle(ROOT);

        // `data-composition-variables` is the single owner of every declared
        // default — parse it rather than repeating each default a second time
        // in JS. `window.__hyperframes` is NOT guaranteed to exist yet when this
        // inline script runs, so it can only ever be an override, never the
        // source. Precedence: host CSS custom property, runtime variable bag,
        // declared default.
        var DECL = {};
        JSON.parse(ROOT.getAttribute("data-composition-variables") || "[]").forEach(function (v) {
          DECL[v.id] = v.default;
        });
        var HF = window.__hyperframes;
        var V = (HF && HF.getVariables && HF.getVariables()) || {};

        function raw(id) {
          var css = CS.getPropertyValue("--" + id.toLowerCase()).trim();
          if (css !== "") return css;
          if (V[id] !== undefined && V[id] !== "") return V[id];
          return DECL[id];
        }
        function num(id) {
          var n = parseFloat(raw(id));
          return isFinite(n) ? n : 0;
        }
        function str(id) {
          var s = raw(id);
          return typeof s === "string" ? s.trim() : "";
        }
        function list(id) {
          return str(id)
            .split(",")
            .map(function (s) {
              return s.trim();
            })
            .filter(function (s) {
              return s !== "";
            });
        }

        var SPEED = num("speed");
        var GRID = Math.max(1, Math.min(4, Math.round(num("grid"))));
        var FILL = num("fill");
        var IMAGE_MIX = num("imageMix");
        var FOG_FAR = Math.max(3, Math.min(14.5, num("fog")));
        var EDGE_OPACITY = num("edgeOpacity");
        var EDGE_COLOR = str("edgeColor");
        var BACKDROP = str("backdrop");
        var PALETTE = list("palette");
        var IMAGE_SRCS = list("images");

        document.getElementById("gt-backdrop").style.background = BACKDROP;

        // How far behind the camera a segment travels before it recycles. Tying
        // it to the fog distance is the whole trick: the recycle boundary lands
        // at exactly FOG_FAR, so a segment reappears already fogged to pure
        // backdrop and nothing ever pops in at the far end. One owner for the
        // invariant — move the fog and the boundary follows.
        var BACK = Math.max(0, Math.min(1.5, NUM_SEGMENTS * SEGMENT_DEPTH - FOG_FAR));

        // Deterministic 2D integer hash — the only source of "randomness".
        // Content at frame N is a pure function of (absolute segment, slot).
        function hash2(a, b) {
          var h = Math.imul(a | 0, 0x27d4eb2d) ^ Math.imul(b | 0, 0x165667b1);
          h = Math.imul(h ^ (h >>> 15), 0x2545f491);
          h ^= h >>> 13;
          return (h >>> 0) / 4294967296;
        }

        var THREEJS = window.THREE;
        var canvas = document.getElementById("gt-canvas");
        var renderer = new THREEJS.WebGLRenderer({
          canvas: canvas,
          antialias: true,
          alpha: false,
          preserveDrawingBuffer: true,
        });
        renderer.setPixelRatio(1);
        renderer.setSize(W, H, false);
        renderer.setClearColor(new THREEJS.Color(BACKDROP), 1);

        var scene = new THREEJS.Scene();
        scene.fog = new THREEJS.Fog(new THREEJS.Color(BACKDROP), 1, FOG_FAR * SEGMENT_DEPTH);

        // Camera sits still at the origin looking down -z; the corridor moves.
        // Same picture as a moving camera, and the modulo keeps every coordinate
        // bounded no matter how long the clip runs.
        var camera = new THREEJS.PerspectiveCamera(FOV, W / H, 0.05, 40);
        camera.position.set(0, 0, 0);
        camera.lookAt(0, 0, -1);

        var PLANE = new THREEJS.PlaneGeometry(1, 1);

        var COLOR_MATS = PALETTE.map(function (hex) {
          return new THREEJS.MeshBasicMaterial({ color: new THREEJS.Color(hex), fog: true });
        });
        if (COLOR_MATS.length === 0) {
          COLOR_MATS.push(new THREEJS.MeshBasicMaterial({ color: 0xffffff, fog: true }));
        }

        // Caller-supplied content slots. Real <img> elements (so the page load
        // event waits for the decode) used directly as WebGL textures.
        var assets = document.getElementById("gt-assets");
        var IMAGES = IMAGE_SRCS.map(function (src, k) {
          var el = document.createElement("img");
          el.id = "gt-img-" + k;
          el.alt = "";
          el.src = src;
          assets.appendChild(el);

          var tex = new THREEJS.Texture(el);
          tex.minFilter = THREEJS.LinearFilter;
          tex.magFilter = THREEJS.LinearFilter;
          tex.generateMipmaps = false;
          var rec = {
            aspect: 0,
            mat: new THREEJS.MeshBasicMaterial({ map: tex, fog: true }),
          };
          function ready() {
            if (!el.naturalWidth) return;
            rec.aspect = el.naturalWidth / el.naturalHeight;
            tex.needsUpdate = true;
          }
          el.addEventListener("load", ready);
          ready(); // a cache hit can complete before the listener is attached
          return rec;
        });

        var edgeMat = new THREEJS.MeshBasicMaterial({
          color: new THREEJS.Color(EDGE_COLOR),
          fog: true,
          transparent: true,
          opacity: EDGE_OPACITY,
        });

        var HW = TUNNEL_WIDTH / 2;
        var HH = TUNNEL_HEIGHT / 2;
        var cellZ = SEGMENT_DEPTH / GRID;
        var cellX = TUNNEL_WIDTH / GRID;
        var cellY = TUNNEL_HEIGHT / GRID;

        // Four faces of a segment box. `place` maps a cell to a world position
        // inside the segment (origin at the segment's near boundary, growing
        // toward -z); `rot` orients the unit plane into that face, inward-facing.
        // `cw`/`ch` are the cell's extent in the plane's own local axes.
        var FACES = [
          {
            // floor
            rot: [-Math.PI / 2, 0, 0],
            cw: cellX,
            ch: cellZ,
            place: function (iu, iz) {
              return [-HW + (iu + 0.5) * cellX, -HH - SKIN, -(iz + 0.5) * cellZ];
            },
          },
          {
            // ceiling
            rot: [Math.PI / 2, 0, 0],
            cw: cellX,
            ch: cellZ,
            place: function (iu, iz) {
              return [-HW + (iu + 0.5) * cellX, HH + SKIN, -(iz + 0.5) * cellZ];
            },
          },
          {
            // left wall
            rot: [0, Math.PI / 2, 0],
            cw: cellZ,
            ch: cellY,
            place: function (iu, iz) {
              return [-HW - SKIN, -HH + (iu + 0.5) * cellY, -(iz + 0.5) * cellZ];
            },
          },
          {
            // right wall
            rot: [0, -Math.PI / 2, 0],
            cw: cellZ,
            ch: cellY,
            place: function (iu, iz) {
              return [HW + SKIN, -HH + (iu + 0.5) * cellY, -(iz + 0.5) * cellZ];
            },
          },
        ];

        var UP = new THREEJS.Vector3(0, 1, 0);
        // Edges are solid tube geometry, not lines: a cylinder thickens with
        // perspective the way a real moulding does, which is the whole reason
        // the corridor reads as built rather than drawn.
        function tube(group, ax, ay, az, bx, by, bz) {
          var dx = bx - ax,
            dy = by - ay,
            dz = bz - az;
          var len = Math.sqrt(dx * dx + dy * dy + dz * dz);
          if (len < 1e-6) return;
          var mesh = new THREEJS.Mesh(
            new THREEJS.CylinderGeometry(LINE_RADIUS, LINE_RADIUS, len, 6, 1),
            edgeMat,
          );
          mesh.position.set((ax + bx) / 2, (ay + by) / 2, (az + bz) / 2);
          mesh.quaternion.setFromUnitVectors(UP, new THREEJS.Vector3(dx / len, dy / len, dz / len));
          group.add(mesh);
        }

        function buildEdges(group) {
          var D = SEGMENT_DEPTH;
          var k, j, u, z;
          // The four corner runs (and any interior wall divisions) come from the
          // side walls; floor/ceiling only contribute their interior lines, so
          // corners are drawn once and never double-blend at 50% opacity.
          for (k = 1; k < GRID; k++) {
            u = -HW + k * cellX;
            tube(group, u, -HH, 0, u, -HH, -D);
            tube(group, u, HH, 0, u, HH, -D);
          }
          for (k = 0; k <= GRID; k++) {
            u = -HH + k * cellY;
            tube(group, -HW, u, 0, -HW, u, -D);
            tube(group, HW, u, 0, HW, u, -D);
          }
          // Ring lines at each depth boundary. The far boundary belongs to the
          // next segment, so stop one short and the tiling stays seamless.
          for (j = 0; j < GRID; j++) {
            z = -j * cellZ;
            tube(group, -HW, -HH, z, HW, -HH, z);
            tube(group, -HW, HH, z, HW, HH, z);
            tube(group, -HW, -HH, z, -HW, HH, z);
            tube(group, HW, -HH, z, HW, HH, z);
          }
        }

        var segs = [];
        for (var i = 0; i < NUM_SEGMENTS; i++) {
          var group = new THREEJS.Group();
          var panels = [];
          for (var f = 0; f < FACES.length; f++) {
            var face = FACES[f];
            for (var iz = 0; iz < GRID; iz++) {
              for (var iu = 0; iu < GRID; iu++) {
                var mesh = new THREEJS.Mesh(PLANE, COLOR_MATS[0]);
                var p = face.place(iu, iz);
                mesh.position.set(p[0], p[1], p[2]);
                mesh.rotation.set(face.rot[0], face.rot[1], face.rot[2]);
                mesh.visible = false;
                group.add(mesh);
                panels.push({ mesh: mesh, cw: face.cw * INSET, ch: face.ch * INSET });
              }
            }
          }
          buildEdges(group);
          scene.add(group);
          segs.push({ group: group, panels: panels });
        }

        // Contain-fit so a supplied screenshot keeps its aspect inside the slot
        // instead of being squashed to the cell (wall cells are portrait).
        function fit(p, ar) {
          if (!ar) {
            p.mesh.scale.set(p.cw, p.ch, 1);
            return;
          }
          if (ar > p.cw / p.ch) p.mesh.scale.set(p.cw, p.cw / ar, 1);
          else p.mesh.scale.set(p.ch * ar, p.ch, 1);
        }

        // Everything about frame t is computed from t alone. `d` is the distance
        // flown in segments; `a` is the ABSOLUTE segment number occupying pool
        // slot i right now, which is what content hashes and the every-other
        // rhythm key off — so a segment carries the same panels every time it is
        // looked at, no matter how the timeline is scrubbed.
        function draw(t) {
          var d = SPEED * t;
          for (var i = 0; i < NUM_SEGMENTS; i++) {
            var seg = segs[i];
            var a = i + NUM_SEGMENTS * Math.ceil((d - i - BACK) / NUM_SEGMENTS);
            seg.group.position.z = -(a - d) * SEGMENT_DEPTH;

            var lit = a % 2 === 0;
            var panels = seg.panels;
            for (var s = 0; s < panels.length; s++) {
              var p = panels[s];
              if (!lit || hash2(a, s * 4 + 1) >= FILL) {
                p.mesh.visible = false;
                continue;
              }
              p.mesh.visible = true;
              if (IMAGES.length > 0 && hash2(a, s * 4 + 2) < IMAGE_MIX) {
                var im = IMAGES[Math.floor(hash2(a, s * 4 + 4) * IMAGES.length) % IMAGES.length];
                p.mesh.material = im.mat;
                fit(p, im.aspect);
              } else {
                var ci = Math.floor(hash2(a, s * 4 + 3) * COLOR_MATS.length) % COLOR_MATS.length;
                p.mesh.material = COLOR_MATS[ci];
                p.mesh.scale.set(p.cw, p.ch, 1);
              }
            }
          }
          renderer.render(scene, camera);
        }

        window.__timelines = window.__timelines || {};
        var tl = gsap.timeline({ paused: true });

        // Repaint from a property SETTER, not from onUpdate: gsap's seek(t)
        // suppresses events by default, so an onUpdate callback silently never
        // fires on a scrub and the canvas freezes on frame 0. Tweened values are
        // always written during render, suppressed or not.
        var driver = { _t: 0 };
        Object.defineProperty(driver, "t", {
          get: function () {
            return this._t;
          },
          set: function (v) {
            this._t = v;
            draw(v);
          },
        });
        tl.to(driver, { t: DUR, duration: DUR, ease: "none" }, 0);
        window.__timelines["gallery-tunnel"] = tl;

        draw(0);
      })();
    </script>
  </body>
</html>
```

Copy npx hyperframes add gallery-tunnel to the clipboard

That writes one file: `compositions/gallery-tunnel.html`.

## Add it to your video

It runs for 12 seconds at 1920×1080. Paste this into your composition:

```html index.html
<div
  data-composition-id="gallery-tunnel"
  data-composition-src="compositions/gallery-tunnel.html"
  data-start="0"
  data-duration="12"
  data-track-index="1"
  data-width="1920"
  data-height="1080"
></div>
```

Move it in time with `data-start`. Put it on a different timeline row with
`data-track-index`. See [data attributes](/concepts/data-attributes.md) for the rest.

## Change how it looks

Set these CSS variables on the block:

* `--images` — Image slots: comma-separated paths, tiled into the panels (e.g. assets/shot-1.png,assets/shot-2.png). Defaults to \`\`.
* `--imagemix` — Image share of filled slots (0 = all colour, 1 = all images). Defaults to `0.5`.
* `--speed` — Flight speed (segments per second). Defaults to `3`.
* `--palette` — Slab palette: comma-separated hex colours. Defaults to `#FF6A00,#AB54F7,#EA3737,#0072E3,#00AA3C,#FFB200`.
* `--grid` — Grid density: slots per wall face per axis (2 = 4 per face, 16 per segment). Defaults to `2`.
* `--fill` — Slot fill rate. Defaults to `0.5`.
* `--fog` — Fog distance in segments: depth at which the corridor reaches pure backdrop. Defaults to `14.25`.
* `--edgecolor` — Edge line colour. Defaults to `#B0B0B0`.
* `--edgeopacity` — Edge line opacity. Defaults to `0.5`.
* `--backdrop` — Backdrop and fog colour. Defaults to `#000000`.

## Variables

Every one of these has a default, so the piece works untouched. Set the ones you
want to change on the element:

| Variable | Default | Accepts | What it does |
| - | - | - | - |
| `images` | \`\` | string | |
| `imageMix` | `0.5` | 0 to 1, step 0.05 | |
| `speed` | `3` | 0segments/s to 8segments/s, step 0.1segments/s | |
| `palette` | `#FF6A00,#AB54F7,#EA3737,#0072E3,#00AA3C,#FFB200` | string | |
| `grid` | `2` | 1 to 4, step 1 | |
| `fill` | `0.5` | 0 to 1, step 0.05 | |
| `fog` | `14.25` | 3segments to 14.5segments, step 0.25segments | |
| `edgeColor` | `#B0B0B0` | color | |
| `edgeOpacity` | `0.5` | 0 to 1, step 0.05 | |
| `backdrop` | `#000000` | color | |

Set them with `data-variable-values` on the element that mounts it. These are the
defaults, so this behaves exactly like the preview above until you change one:

```html wrap
<div
  data-composition-id="gallery-tunnel"
  data-composition-src="compositions/gallery-tunnel.html"
  data-variable-values='{"images":"","imageMix":0.5,"speed":3,"palette":"#FF6A00,#AB54F7,#EA3737,#0072E3,#00AA3C,#FFB200","grid":2,"fill":0.5,"fog":14.25,"edgeColor":"#B0B0B0","edgeOpacity":0.5,"backdrop":"#000000"}'
></div>
```

{/* hf:generated-footer */}

Tagged `webgl` `3d` `background` `showcase` `depth` `loop`.

## Related topics

* [Browse the complete Catalog](/catalog.md)
* [Add assets and Catalog items in the browser editor](/studio/assets-and-blocks.md)
* [Put blocks together into one video](/concepts/compositions.md)
