feat(cnn_v3): GBufferEffect internal scene + GBufViewEffect debug wiring

GBufferEffect:
- set_scene() now owns Scene/Camera internally; no external pointers needed
- 20 randomly rotating cubes (xorshift32 seed, axis-angle animation)
- 4 pumping spheres (radius = base_r * (1 + audio_intensity * 0.8))
- Camera at (0,2.5,6) looking at origin; aspect updated per-frame
- GBufLightsUniforms: 2 directional lights (warm key + cool fill)
- object_type written to ObjectData.params.x (ready for SDF shadow)
- shadow/transp nodes cleared via zero-draw render passes (placeholder)
- bilinear sampler cached via create_linear_sampler() / sampler_.get()
- dead placeholder textures removed

GBufViewEffect:
- gbuf_view.wgsl: all channels now fully grayscale (removed color tint)
- seq_compiler.py: GBufViewEffect added to CLASS_TO_HEADER
- timeline.seq: cnn_v3_test uses GBufViewEffect -> sink for debug view

Docs: HOWTO.md §1 updated with set_scene() description + §1b implementation
plan for Pass 2 SDF shadow (shader spec, bind layout, C++ additions)

handoff(Gemini): GBufferEffect has internal scene, 36/36 tests green.
Next: implement Pass 2 shadow (gbuf_shadow.wgsl) per §1b plan in HOWTO.md.

6 files changed, 279 insertions, 119 deletions

diff --git a/cnn_v3/docs/HOWTO.md b/cnn_v3/docs/HOWTO.md
index 08979e7..2d88019 100644
--- a/cnn_v3/docs/HOWTO.md
+++ b/cnn_v3/docs/HOWTO.md
@@ -22,57 +22,141 @@ It rasterizes proxy geometry to MRT G-buffer textures and packs them into two
 
 ### Adding to a Sequence
 
-`GBufferEffect` does not exist in `seq_compiler.py` as a named effect yet
-(no `.seq` syntax integration for Phase 1). Wire it directly in C++ alongside
-your scene code, or add it to the timeline when the full CNNv3Effect is ready.
+Both `GBufferEffect` and `GBufViewEffect` are registered in `seq_compiler.py`
+(`CLASS_TO_HEADER`) and can be wired directly in `timeline.seq`.
 
-**C++ wiring example** (e.g. inside a Sequence or main.cc):
+**Debug view (G-buffer → sink)**:
+```seq
+SEQUENCE 12.00 0 "cnn_v3_test"
+  NODE gbuf_feat0 gbuf_rgba32uint
+  NODE gbuf_feat1 gbuf_rgba32uint
+  EFFECT + GBufferEffect source -> gbuf_feat0 gbuf_feat1 0.00 8.00
+  EFFECT + GBufViewEffect gbuf_feat0 gbuf_feat1 -> sink 0.00 8.00
+```
 
-```cpp
-#include "../../cnn_v3/src/gbuffer_effect.h"
+**Full CNN pipeline**:
+```seq
+SEQUENCE 12.00 0 "cnn_v3_test"
+  NODE gbuf_feat0 gbuf_rgba32uint
+  NODE gbuf_feat1 gbuf_rgba32uint
+  NODE cnn_v3_out gbuf_albedo
+  EFFECT + GBufferEffect source -> gbuf_feat0 gbuf_feat1 0.00 8.00
+  EFFECT + CNNv3Effect gbuf_feat0 gbuf_feat1 -> cnn_v3_out 0.00 8.00
+  EFFECT + Passthrough cnn_v3_out -> sink 0.00 8.00
+```
 
-// Allocate once alongside your scene
-auto gbuf = std::make_shared<GBufferEffect>(
-    ctx, /*inputs=*/{"prev_cnn"},  // or any dummy node
-    /*outputs=*/{"gbuf_feat0", "gbuf_feat1"},
-    /*start=*/0.0f, /*end=*/60.0f);
+### Internal scene
 
-gbuf->set_scene(&my_scene, &my_camera);
+Call `set_scene()` once before the first render to populate the built-in demo
+scene. No external `Scene` or `Camera` pointer is required — the effect owns
+them.
 
-// In render loop, call before CNN pass:
-gbuf->render(encoder, params, nodes);
-```
+**What `set_scene()` creates:**
+- **20 small cubes** — random positions in [-2,2]×[-1.5,1.5]³, scale 0.1–0.25,
+  random colors. Each has a random rotation axis and speed; animated each frame
+  via `quat::from_axis(axis, time * speed)`.
+- **4 pumping spheres** — at fixed world positions, base radii 0.25–0.35.
+  Scale driven by `audio_intensity`: `r = base_r * (1 + audio_intensity * 0.8)`.
+- **Camera** — position (0, 2.5, 6), target (0, 0, 0), 45° FOV.
+  Aspect ratio updated each frame from `params.aspect_ratio`.
+- **Two directional lights** (uploaded to `lights_uniform_`, ready for shadow pass):
+  - Key: warm white (1.0, 0.92, 0.78), direction `normalize(1, 2, 1)` (upper-right-front)
+  - Fill: cool blue (0.4, 0.45, 0.8 × 0.4), direction `normalize(-1, 1, -1)` (upper-left-back)
 
 ### Internal passes
 
 Each frame, `GBufferEffect::render()` executes:
 
-1. **Pass 1 — MRT rasterization** (`gbuf_raster.wgsl`)
+1. **Pass 1 — MRT rasterization** (`gbuf_raster.wgsl`) ✅
    - Proxy box (36 verts) × N objects, instanced
    - MRT outputs: `gbuf_albedo` (rgba16float), `gbuf_normal_mat` (rgba16float)
    - Depth test + write into `gbuf_depth` (depth32float)
+   - `obj.type` written to `ObjectData.params.x` for future SDF branching
+
+2. **Pass 2 — SDF shadow raymarching** (`gbuf_shadow.wgsl`) — TODO
+   - See implementation plan below.
 
-2. **Pass 2/3 — SDF + Lighting** — TODO (placeholder: shadow=1, transp=0)
+3. **Pass 3 — Transparency** — TODO (deferred; transp=0 for opaque scenes)
 
-3. **Pass 4 — Pack compute** (`gbuf_pack.wgsl`)
+4. **Pass 4 — Pack compute** (`gbuf_pack.wgsl`) ✅
    - Reads all G-buffer textures + `prev_cnn` input
    - Writes `feat_tex0` + `feat_tex1` (rgba32uint, 20 channels, 32 bytes/pixel)
+   - Shadow / transp nodes cleared to 1.0 / 0.0 via zero-draw render passes
+     until Pass 2/3 are implemented.
 
 ### Output node names
 
-By default the outputs are named from the `outputs` vector passed to the
-constructor. Use these names when binding the CNN effect input:
+Outputs are named from the `outputs` vector passed to the constructor:
 
 ```
 outputs[0]  → feat_tex0   (rgba32uint: albedo.rgb, normal.xy, depth, depth_grad.xy)
 outputs[1]  → feat_tex1   (rgba32uint: mat_id, prev.rgb, mip1.rgb, mip2.rgb, shadow, transp)
 ```
 
-### Scene data
+---
+
+## 1b. GBufferEffect — Implementation Plan (Pass 2: SDF Shadow)
+
+### What remains
+
+| Item | Status | Notes |
+|------|--------|-------|
+| Pass 1: MRT raster | ✅ Done | proxy box, all object types |
+| Pass 4: Pack compute | ✅ Done | 20 channels packed |
+| Internal scene + animation | ✅ Done | cubes + spheres + 2 lights |
+| Pass 2: SDF shadow | ❌ TODO | main missing piece |
+| Pass 3: Transparency | ❌ TODO | low priority, opaque scenes only |
+| Phase 4: type-aware SDF | ❌ TODO | optional refinement |
+
+### Pass 2: SDF shadow raymarching
+
+**New file: `cnn_v3/shaders/gbuf_shadow.wgsl`** — fullscreen render pass.
+
+Bind layout:
+
+| Binding | Type | Content |
+|---------|------|---------|
+| 0 | `uniform` | `GlobalUniforms` (`#include "common_uniforms"`) |
+| 1 | `storage read` | `ObjectsBuffer` |
+| 2 | `texture_depth_2d` | depth from Pass 1 |
+| 3 | `sampler` (non-filtering) | depth load |
+| 4 | `uniform` | `GBufLightsUniforms` (2 lights) |
+
+Algorithm per fragment:
+1. Reconstruct world position from NDC depth + `globals.inv_view_proj`
+2. For each object: `sdBox((inv_model * world_pos).xyz, vec3(1.0))` — proxy box in local space
+3. For each light: offset ray origin by `0.02 * surface_normal`; march shadow ray toward `light.direction`
+4. Soft shadow via `shadowWithStoredDistance()` from `render/raymarching_id`
+5. Combine lights: `shadow = min(shadow_light0, shadow_light1)`
+6. Discard fragments where depth == 1.0 (sky/background → shadow = 1.0)
+7. Output shadow factor to RGBA8Unorm render target (`.r` = shadow)
+
+**C++ additions (`gbuffer_effect.h/.cc`):**
+```cpp
+RenderPipeline shadow_pipeline_;
+void create_shadow_pipeline();
+```
+In `render()` between Pass 1 and the shadow/transp node clears:
+- Build bind group (global_uniforms_buf_, objects_buf_, depth_view, sampler_, lights_uniform_)
+- Run fullscreen triangle → `node_shadow_` color attachment
+- Remove the `clear_node(node_shadow_, 1.0f)` placeholder once the pass is live
+
+**Register:**
+- `cnn_v3/shaders/gbuf_shadow.wgsl` → `SHADER_GBUF_SHADOW` in `assets.txt`
+- `extern const char* gbuf_shadow_wgsl;` in `gbuffer_effect.cc`
+
+### Phase 4: Object-type-aware SDF (optional)
+
+Branch on `obj.params.x` (populated since this commit) using `math/sdf_shapes`:
+
+| Type value | ObjectType | SDF |
+|------------|-----------|-----|
+| 0 | CUBE | `sdBox(local_p, vec3(1))` |
+| 1 | SPHERE | `sdSphere(local_p, 1.0)` |
+| 2 | PLANE | `sdPlane(local_p, vec3(0,1,0), obj.params.y)` |
+| 3 | TORUS | `sdTorus(local_p, vec2(0.8, 0.2))` |
 
-Call `set_scene(scene, camera)` before the first render. The effect uploads
-`GlobalUniforms` (view-proj, camera pos, resolution) and `ObjectData` (model
-matrix, color) to GPU storage buffers each frame.
+Only worth adding after Pass 2 is validated visually.
 
 ---
 
@@ -386,13 +470,13 @@ auto gview = std::make_shared<GBufViewEffect>(ctx,
 
 | Row | Col 0 | Col 1 | Col 2 | Col 3 |
 |-----|-------|-------|-------|-------|
-| 0 | `alb.r` (red tint) | `alb.g` (green tint) | `alb.b` (blue tint) | `nrm.x` remap→[0,1] |
+| 0 | `alb.r` | `alb.g` | `alb.b` | `nrm.x` remap→[0,1] |
 | 1 | `nrm.y` remap→[0,1] | `depth` (inverted) | `dzdx` ×20+0.5 | `dzdy` ×20+0.5 |
 | 2 | `mat_id` | `prev.r` | `prev.g` | `prev.b` |
 | 3 | `mip1.r` | `mip1.g` | `mip1.b` | `mip2.r` |
 | 4 | `mip2.g` | `mip2.b` | `shadow` | `transp` |
 
-1-pixel gray grid lines separate cells. Dark background for out-of-range cells.
+All channels displayed as grayscale. 1-pixel gray grid lines separate cells. Dark background for out-of-range cells.
 
 **Shader binding layout** (no sampler needed — integer texture):
 
diff --git a/cnn_v3/shaders/gbuf_view.wgsl b/cnn_v3/shaders/gbuf_view.wgsl
index f2ae085..a5e6c91 100644
--- a/cnn_v3/shaders/gbuf_view.wgsl
+++ b/cnn_v3/shaders/gbuf_view.wgsl
@@ -93,10 +93,5 @@ fn fs_main(@builtin(position) pos: vec4f) -> @location(0) vec4f {
         disp = clamp(v, 0.0, 1.0);
     }
 
-    // Albedo channels: tint for identification (ch0=red, ch1=green, ch2=blue)
-    if      (ch == 0u) { return vec4f(disp, 0.0,  0.0,  1.0); }
-    else if (ch == 1u) { return vec4f(0.0,  disp, 0.0,  1.0); }
-    else if (ch == 2u) { return vec4f(0.0,  0.0,  disp, 1.0); }
-
     return vec4f(disp, disp, disp, 1.0);
 }
diff --git a/cnn_v3/src/gbuffer_effect.cc b/cnn_v3/src/gbuffer_effect.cc
index 750188f..f529d2b 100644
--- a/cnn_v3/src/gbuffer_effect.cc
+++ b/cnn_v3/src/gbuffer_effect.cc
@@ -41,41 +41,6 @@ struct GBufGlobalUniforms {
 static_assert(sizeof(GBufGlobalUniforms) == sizeof(float) * 44,
               "GBufGlobalUniforms must be 176 bytes");
 
-// Helper: create a 1×1 placeholder texture of a given format cleared to `value`.
-static WGPUTexture create_placeholder_tex(WGPUDevice device,
-                                          WGPUTextureFormat format,
-                                          float value) {
-  WGPUTextureDescriptor desc = {};
-  desc.usage = (WGPUTextureUsage)(WGPUTextureUsage_TextureBinding |
-                                  WGPUTextureUsage_CopyDst);
-  desc.dimension = WGPUTextureDimension_2D;
-  desc.size = {1, 1, 1};
-  desc.format = format;
-  desc.mipLevelCount = 1;
-  desc.sampleCount = 1;
-  WGPUTexture tex = wgpuDeviceCreateTexture(device, &desc);
-  return tex;
-}
-
-// Helper: write a single RGBA float pixel to a texture via queue.
-static void write_placeholder_pixel(WGPUQueue queue, WGPUTexture tex,
-                                    float r, float g, float b, float a) {
-  const float data[4] = {r, g, b, a};
-  WGPUTexelCopyTextureInfo dst = {};
-  dst.texture = tex;
-  dst.mipLevel = 0;
-  dst.origin = {0, 0, 0};
-  dst.aspect = WGPUTextureAspect_All;
-
-  WGPUTexelCopyBufferLayout layout = {};
-  layout.offset = 0;
-  layout.bytesPerRow = 16; // 4 × sizeof(float)
-  layout.rowsPerImage = 1;
-
-  const WGPUExtent3D extent = {1, 1, 1};
-  wgpuQueueWriteTexture(queue, &dst, data, sizeof(data), &layout, &extent);
-}
-
 // Create bilinear sampler.
 static WGPUSampler create_bilinear_sampler(WGPUDevice device) {
   WGPUSamplerDescriptor desc = {};
@@ -116,31 +81,9 @@ GBufferEffect::GBufferEffect(const GpuContext& ctx,
 
   // Resolution uniform for pack shader.
   pack_res_uniform_.init(ctx_.device);
+  lights_uniform_.init(ctx_.device);
 
-  // Placeholder shadow (1.0 = fully lit) and transp (0.0 = opaque) textures.
-  shadow_placeholder_tex_.set(
-      create_placeholder_tex(ctx_.device, WGPUTextureFormat_RGBA32Float, 1.0f));
-  write_placeholder_pixel(ctx_.queue,
-                          shadow_placeholder_tex_.get(), 1.0f, 0.0f, 0.0f, 1.0f);
-
-  transp_placeholder_tex_.set(
-      create_placeholder_tex(ctx_.device, WGPUTextureFormat_RGBA32Float, 0.0f));
-  write_placeholder_pixel(ctx_.queue,
-                          transp_placeholder_tex_.get(), 0.0f, 0.0f, 0.0f, 1.0f);
-
-  WGPUTextureViewDescriptor vd = {};
-  vd.format = WGPUTextureFormat_RGBA32Float;
-  vd.dimension = WGPUTextureViewDimension_2D;
-  vd.baseMipLevel = 0;
-  vd.mipLevelCount = 1;
-  vd.baseArrayLayer = 0;
-  vd.arrayLayerCount = 1;
-  vd.aspect = WGPUTextureAspect_All;
-
-  shadow_placeholder_view_.set(
-      wgpuTextureCreateView(shadow_placeholder_tex_.get(), &vd));
-  transp_placeholder_view_.set(
-      wgpuTextureCreateView(transp_placeholder_tex_.get(), &vd));
+  create_linear_sampler();
 
   create_raster_pipeline();
   create_pack_pipeline();
@@ -162,19 +105,113 @@ void GBufferEffect::declare_nodes(NodeRegistry& registry) {
   }
 }
 
-void GBufferEffect::set_scene(const Scene* scene, const Camera* camera) {
-  scene_  = scene;
-  camera_ = camera;
+void GBufferEffect::set_scene() {
+  scene_.clear();
+  cube_anims_.clear();
+  sphere_anims_.clear();
+
+  // Deterministic pseudo-random (xorshift32).
+  uint32_t seed = 0xBEEF1234u;
+  auto rnd = [&]() -> float {
+    seed ^= seed << 13;
+    seed ^= seed >> 17;
+    seed ^= seed << 5;
+    return (float)(seed >> 8) / 16777216.0f;  // [0, 1)
+  };
+  auto rrange = [&](float lo, float hi) { return lo + rnd() * (hi - lo); };
+
+  // 20 small cubes scattered in a [-2,2]×[-1.5,1.5]×[-1.5,1.5] volume.
+  static const int kNumCubes = 20;
+  for (int i = 0; i < kNumCubes; ++i) {
+    Object3D obj(ObjectType::CUBE);
+    obj.position = vec3(rrange(-2.0f, 2.0f),
+                        rrange(-1.5f, 1.5f),
+                        rrange(-1.5f, 1.5f));
+    const float s = rrange(0.10f, 0.25f);
+    obj.scale = vec3(s, s, s);
+    obj.color = vec4(rrange(0.4f, 1.0f),
+                     rrange(0.4f, 1.0f),
+                     rrange(0.4f, 1.0f), 1.0f);
+
+    // Random rotation axis (avoid degenerate zero-length axis).
+    vec3 axis = vec3(rrange(-1.0f, 1.0f),
+                     rrange(-1.0f, 1.0f),
+                     rrange(-1.0f, 1.0f));
+    if (axis.len() < 0.01f) axis = vec3(0.0f, 1.0f, 0.0f);
+    axis = axis.normalize();
+    const float speed = rrange(0.3f, 1.5f) * (rnd() > 0.5f ? 1.0f : -1.0f);
+
+    scene_.add_object(obj);
+    cube_anims_.push_back({axis, speed});
+  }
+
+  // 4 pumping spheres at fixed positions; radius modulated by audio_intensity.
+  static const vec3 kSpherePos[4] = {
+    { 0.0f,  0.0f,  0.0f},
+    { 1.5f,  0.5f, -0.5f},
+    {-1.5f, -0.5f,  0.5f},
+    { 0.0f,  1.0f,  1.0f},
+  };
+  static const float kBaseSphereRadius[4] = {0.35f, 0.28f, 0.30f, 0.25f};
+  for (int i = 0; i < 4; ++i) {
+    Object3D obj(ObjectType::SPHERE);
+    obj.position = kSpherePos[i];
+    const float r = kBaseSphereRadius[i];
+    obj.scale = vec3(r, r, r);
+    obj.color = vec4(0.85f, 0.60f, 0.95f, 1.0f);
+    const int idx = (int)scene_.objects.size();
+    scene_.add_object(obj);
+    sphere_anims_.push_back({idx, r});
+  }
+
+  // Camera: above and in front of the scene, looking at origin.
+  camera_.set_look_at(vec3(0.0f, 2.5f, 6.0f),
+                      vec3(0.0f, 0.0f, 0.0f),
+                      vec3(0.0f, 1.0f, 0.0f));
+  camera_.fov_y_rad  = 0.7854f;  // 45°
+  camera_.near_plane = 0.1f;
+  camera_.far_plane  = 20.0f;
+  // aspect_ratio is updated each frame from params.resolution.
+
+  scene_ready_ = true;
 }
 
 void GBufferEffect::render(WGPUCommandEncoder encoder,
                            const UniformsSequenceParams& params,
                            NodeRegistry& nodes) {
-  if (!scene_ || !camera_) {
+  if (!scene_ready_) {
     return;
   }
 
-  upload_scene_data(*scene_, *camera_, params.time);
+  // Update camera aspect ratio from current resolution.
+  camera_.aspect_ratio = params.aspect_ratio;
+
+  // Animate cubes: axis-angle rotation driven by physical time.
+  for (int i = 0; i < (int)cube_anims_.size(); ++i) {
+    const CubeAnim& a = cube_anims_[(size_t)i];
+    scene_.objects[(size_t)i].rotation =
+        quat::from_axis(a.axis, params.time * a.speed);
+  }
+  // Pump spheres: scale with audio_intensity.
+  for (const SphereAnim& a : sphere_anims_) {
+    const float r = a.base_radius * (1.0f + params.audio_intensity * 0.8f);
+    scene_.objects[(size_t)a.obj_idx].scale = vec3(r, r, r);
+  }
+
+  // Upload two directional lights.
+  {
+    GBufLightsUniforms lu = {};
+    lu.params = vec4(2.0f, 0.0f, 0.0f, 0.0f);
+    // Key: warm sun, upper-right-front.
+    lu.lights[0].direction = vec4(0.408f, 0.816f, 0.408f, 0.0f);  // norm(1,2,1)
+    lu.lights[0].color     = vec4(1.00f, 0.92f, 0.78f, 1.0f);
+    // Fill: cool sky, upper-left-back.
+    lu.lights[1].direction = vec4(-0.577f, 0.577f, -0.577f, 0.0f);  // norm(-1,1,-1)
+    lu.lights[1].color     = vec4(0.40f, 0.45f, 0.80f, 0.4f);
+    lights_uniform_.update(ctx_.queue, lu);
+  }
+
+  upload_scene_data(scene_, camera_, params.time);
 
   // Update resolution uniform for pack shader.
   GBufResUniforms res_uni;
@@ -228,8 +265,8 @@ void GBufferEffect::render(WGPUCommandEncoder encoder,
   raster_pass_desc.depthStencilAttachment = &depth_attachment;
 
   const int num_objects =
-      (int)(scene_->objects.size() < (size_t)kGBufMaxObjects
-                ? scene_->objects.size()
+      (int)(scene_.objects.size() < (size_t)kGBufMaxObjects
+                ? scene_.objects.size()
                 : (size_t)kGBufMaxObjects);
 
   if (num_objects > 0 && raster_pipeline_.get() != nullptr) {
@@ -250,13 +287,31 @@ void GBufferEffect::render(WGPUCommandEncoder encoder,
     wgpuRenderPassEncoderRelease(raster_pass);
   }
 
-  // Pass 2: SDF raymarching — TODO (placeholder: shadow=1, transp=0 already set)
+  // Pass 2: SDF raymarching — TODO
   // Pass 3: Lighting/shadow — TODO
 
+  // Clear shadow node to 1.0 (fully lit) and transp to 0.0 (fully opaque)
+  // until passes 2-3 are implemented.
+  auto clear_node = [&](const std::string& name, float value) {
+    WGPURenderPassColorAttachment att = {};
+    att.view       = nodes.get_view(name);
+    att.loadOp     = WGPULoadOp_Clear;
+    att.storeOp    = WGPUStoreOp_Store;
+    att.clearValue = {value, value, value, value};
+    att.depthSlice = WGPU_DEPTH_SLICE_UNDEFINED;
+    WGPURenderPassDescriptor pd = {};
+    pd.colorAttachmentCount = 1;
+    pd.colorAttachments     = &att;
+    WGPURenderPassEncoder p = wgpuCommandEncoderBeginRenderPass(encoder, &pd);
+    wgpuRenderPassEncoderEnd(p);
+    wgpuRenderPassEncoderRelease(p);
+  };
+  clear_node(node_shadow_, 1.0f);
+  clear_node(node_transp_, 0.0f);
+
   // --- Pass 4: Pack compute ---
   // Rebuild pack bind group with current node views.
-  // Construct a temporary bilinear sampler for this pass.
-  WGPUSampler bilinear = create_bilinear_sampler(ctx_.device);
+  WGPUSampler bilinear = sampler_.get();
 
   // Get texture views from nodes.
   // shadow / transp are GBUF_R8 nodes; use their views.
@@ -320,7 +375,7 @@ void GBufferEffect::render(WGPUCommandEncoder encoder,
   wgpuComputePassEncoderRelease(compute_pass);
 
   wgpuBindGroupRelease(pack_bg);
-  wgpuSamplerRelease(bilinear);
+  // bilinear is owned by sampler_ — no release here.
 }
 
 // ---- private helpers ----
@@ -373,7 +428,7 @@ void GBufferEffect::upload_scene_data(const Scene& scene,
       d.model     = m;
       d.inv_model = m.inverse();
       d.color     = obj.color;
-      d.params    = vec4(0.0f, 0.0f, 0.0f, 0.0f);
+      d.params    = vec4((float)(int)obj.type, 0.0f, 0.0f, 0.0f);
       obj_data.push_back(d);
     }
     wgpuQueueWriteBuffer(ctx_.queue, objects_buf_.buffer, 0,
diff --git a/cnn_v3/src/gbuffer_effect.h b/cnn_v3/src/gbuffer_effect.h
index 42fb0ec..d45be75 100644
--- a/cnn_v3/src/gbuffer_effect.h
+++ b/cnn_v3/src/gbuffer_effect.h
@@ -10,6 +10,7 @@
 #include "gpu/uniform_helper.h"
 #include "gpu/wgpu_resource.h"
 #include "util/mini_math.h"
+#include <vector>
 
 // Uniform for the pack compute shader
 struct GBufResUniforms {
@@ -20,6 +21,20 @@ struct GBufResUniforms {
 static_assert(sizeof(GBufResUniforms) == 16,
               "GBufResUniforms must be 16 bytes");
 
+// Single directional light: direction points *toward* the light source (world space).
+struct GBufLight {
+  vec4 direction;  // xyz = normalized direction toward light, w = unused
+  vec4 color;      // rgb = color, a = intensity
+};
+static_assert(sizeof(GBufLight) == 32, "GBufLight must be 32 bytes");
+
+struct GBufLightsUniforms {
+  GBufLight lights[2];
+  vec4      params;  // x = num_lights
+};
+static_assert(sizeof(GBufLightsUniforms) == 80,
+              "GBufLightsUniforms must be 80 bytes");
+
 class GBufferEffect : public Effect {
  public:
   GBufferEffect(const GpuContext& ctx, const std::vector<std::string>& inputs,
@@ -31,9 +46,22 @@ class GBufferEffect : public Effect {
   void render(WGPUCommandEncoder encoder, const UniformsSequenceParams& params,
               NodeRegistry& nodes) override;
 
-  void set_scene(const Scene* scene, const Camera* camera);
+  // Populate the internal scene with ~20 rotating cubes and a few pumping
+  // spheres.  Must be called once before the first render().
+  void set_scene();
 
  private:
+  // Per-cube animation state (axis-angle rotation)
+  struct CubeAnim {
+    vec3  axis;
+    float speed;  // radians/second, may be negative
+  };
+  // Per-sphere animation state (radius driven by audio_intensity)
+  struct SphereAnim {
+    int   obj_idx;     // index into scene_.objects
+    float base_radius;
+  };
+
   // Internal G-buffer node names
   std::string node_albedo_;
   std::string node_normal_mat_;
@@ -43,8 +71,13 @@ class GBufferEffect : public Effect {
   std::string node_feat0_;
   std::string node_feat1_;
 
-  const Scene*  scene_  = nullptr;
-  const Camera* camera_ = nullptr;
+  // Owned scene and camera — populated by set_scene()
+  Scene  scene_;
+  Camera camera_;
+  bool   scene_ready_ = false;
+
+  std::vector<CubeAnim>  cube_anims_;
+  std::vector<SphereAnim> sphere_anims_;
 
   // Pass 1: MRT rasterization pipeline
   RenderPipeline raster_pipeline_;
@@ -53,19 +86,13 @@ class GBufferEffect : public Effect {
   // Pass 4: Pack compute pipeline
   ComputePipeline  pack_pipeline_;
   BindGroup        pack_bind_group_;
-  UniformBuffer<GBufResUniforms> pack_res_uniform_;
-
-  // Placeholder textures for shadow/transp (white/black cleared once)
-  Texture     shadow_placeholder_tex_;
-  TextureView shadow_placeholder_view_;
-  Texture     transp_placeholder_tex_;
-  TextureView transp_placeholder_view_;
+  UniformBuffer<GBufResUniforms>    pack_res_uniform_;
+  UniformBuffer<GBufLightsUniforms> lights_uniform_;
 
   // GPU-side object data buffers (global uniforms + objects storage)
-  // These mirror the layout expected by gbuf_raster.wgsl
   GpuBuffer global_uniforms_buf_;
   GpuBuffer objects_buf_;
-  int       objects_buf_capacity_ = 0; // number of ObjectData slots allocated
+  int       objects_buf_capacity_ = 0;
 
   void create_raster_pipeline();
   void create_pack_pipeline();
diff --git a/tools/seq_compiler.py b/tools/seq_compiler.py
index 09188a5..2d802b2 100755
--- a/tools/seq_compiler.py
+++ b/tools/seq_compiler.py
@@ -397,8 +397,9 @@ def generate_cpp(seq: SequenceDecl, sorted_effects: List[EffectDecl],
     # Use a full #include string to override the path entirely.
     CLASS_TO_HEADER = {
         'NtscYiq':       'ntsc',
-        'GBufferEffect': '#include "../../cnn_v3/src/gbuffer_effect.h"',
-        'CNNv3Effect':   '#include "../../cnn_v3/src/cnn_v3_effect.h"',
+        'GBufferEffect':  '#include "../../cnn_v3/src/gbuffer_effect.h"',
+        'CNNv3Effect':    '#include "../../cnn_v3/src/cnn_v3_effect.h"',
+        'GBufViewEffect': '#include "../../cnn_v3/src/gbuf_view_effect.h"',
     }
     includes = set()
     for effect in seq.effects:
diff --git a/workspaces/main/timeline.seq b/workspaces/main/timeline.seq
index 557e73d..0e1ca74 100644
--- a/workspaces/main/timeline.seq
+++ b/workspaces/main/timeline.seq
@@ -13,10 +13,8 @@ SEQUENCE 8.00 0 "scene 2"
 SEQUENCE 12.00 0 "cnn_v3_test"
   NODE gbuf_feat0 gbuf_rgba32uint
   NODE gbuf_feat1 gbuf_rgba32uint
-  NODE cnn_v3_out gbuf_albedo
   EFFECT + GBufferEffect source -> gbuf_feat0 gbuf_feat1 0.00 8.00
-  EFFECT + CNNv3Effect gbuf_feat0 gbuf_feat1 -> cnn_v3_out 0.00 8.00
-  EFFECT + Passthrough cnn_v3_out -> sink 0.00 8.00
+  EFFECT + GBufViewEffect gbuf_feat0 gbuf_feat1 -> sink 0.00 8.00
 
 SEQUENCE 20.00 2 "hybrid_heptagon"
   # Heptagon -> Hybrid3D -> sink