diff --git a/src/bin/minitracer/main.rs b/src/bin/minitracer/main.rs
index 7faf46e..73cf45d 100644
--- a/src/bin/minitracer/main.rs
+++ b/src/bin/minitracer/main.rs
@@ -1,6 +1,7 @@
 use std::error::Error;
 
-use glam::uvec2;
+use glam::{uvec2, vec2};
+use raytracing3::glare::Glarer;
 use raytracing3::present::{self, Presenter};
 use raytracing3::scene::{Renderer, SceneParams};
 use winit::{
@@ -27,6 +28,7 @@ fn main() {
 	let hdr_format = wgpu::TextureFormat::Rgba16Float;
 	let scene = SceneParams::new(N_SPHERES);
 	let renderer = Renderer::new(&device);
+	let glarer = Glarer::new(&device, hdr_format);
 	let presenter = Presenter::new(&device, output_format);
 
 	let mut frame = 0;
@@ -72,13 +74,13 @@ fn main() {
 						usage: wgpu::TextureUsages::RENDER_ATTACHMENT | wgpu::TextureUsages::TEXTURE_BINDING,
 						view_formats: &[],
 					});
-					let hdr = hdr.create_view(&wgpu::TextureViewDescriptor::default());
+					let hdr1v = hdr.create_view(&wgpu::TextureViewDescriptor::default());
 					let mut encoder = device.create_command_encoder(&wgpu::CommandEncoderDescriptor { label: None });
 					{
-						let mut render_pass = renderer.prepare(&mut encoder, &hdr);
+						let mut render_pass = renderer.prepare(&mut encoder, &hdr1v);
+						let time = frame as f32 / (60. * RAYS_PER_PIXEL as f32);
 						for _ in 0..RAYS_PER_PIXEL {
 							frame += 1;
-							let time = frame as f32 / (60. * RAYS_PER_PIXEL as f32);
 							renderer.render_frame(
 								&device,
 								&mut render_pass,
@@ -89,10 +91,12 @@ fn main() {
 							);
 						}
 					}
+					let sigma = vec2(size.width as f32, size.height as f32).length();
+					glarer.render(&device, &mut encoder, &hdr, raytracing3::glare::Params { sigma });
 					presenter.render(
 						&device,
 						&mut encoder,
-						&hdr,
+						&hdr1v,
 						&view,
 						present::Params {
 							divisor: RAYS_PER_PIXEL as f32,
diff --git a/src/glare.rs b/src/glare.rs
new file mode 100644
index 0000000..7d6882a
--- /dev/null
+++ b/src/glare.rs
@@ -0,0 +1,222 @@
+use std::mem::offset_of;
+
+use bytemuck::{bytes_of, Pod, Zeroable};
+use glam::{vec2, UVec2, Vec2};
+use wgpu::util::DeviceExt;
+
+#[derive(Debug, Clone, Copy, Pod, Zeroable)]
+#[repr(C)]
+struct Vertex {
+	pub screen: Vec2,
+}
+
+#[derive(Debug, Clone, Copy, Pod, Zeroable)]
+#[repr(C)]
+pub struct Params {
+	pub sigma: f32, // in pixels
+}
+
+pub struct Glarer {
+	view_buf: wgpu::Buffer,
+	format: wgpu::TextureFormat,
+	pipeline_h: wgpu::RenderPipeline,
+	pipeline_v: wgpu::RenderPipeline,
+}
+
+static SHADER: &str = include_str!("glare.wgsl");
+
+impl Glarer {
+	pub fn new(device: &wgpu::Device, format: wgpu::TextureFormat) -> Self {
+		let view_buf = device.create_buffer_init(&wgpu::util::BufferInitDescriptor {
+			label: None,
+			contents: bytes_of(&[(0., 0.), (1., 0.), (0., 1.), (1., 1.)].map(|(x, y)| Vertex { screen: vec2(x, y) })),
+			usage: wgpu::BufferUsages::VERTEX,
+		});
+		let shader = device.create_shader_module(wgpu::ShaderModuleDescriptor {
+			label: None,
+			source: wgpu::ShaderSource::Wgsl(SHADER.into()),
+		});
+		let vertex_state = wgpu::VertexState {
+			module: &shader,
+			entry_point: None,
+			compilation_options: wgpu::PipelineCompilationOptions::default(),
+			buffers: &[wgpu::VertexBufferLayout {
+				array_stride: size_of::<Vertex>() as u64,
+				step_mode: wgpu::VertexStepMode::Vertex,
+				attributes: &[wgpu::VertexAttribute {
+					shader_location: 0,
+					offset: offset_of!(Vertex, screen) as u64,
+					format: wgpu::VertexFormat::Float32x2,
+				}],
+			}],
+		};
+		let pipeline1 = device.create_render_pipeline(&wgpu::RenderPipelineDescriptor {
+			label: None,
+			layout: None,
+			vertex: vertex_state.clone(),
+			primitive: wgpu::PrimitiveState {
+				topology: wgpu::PrimitiveTopology::TriangleStrip,
+				..Default::default()
+			},
+			depth_stencil: None,
+			multisample: wgpu::MultisampleState::default(),
+			fragment: Some(wgpu::FragmentState {
+				module: &shader,
+				entry_point: Some("pass_h"),
+				compilation_options: wgpu::PipelineCompilationOptions::default(),
+				targets: &[Some(wgpu::ColorTargetState {
+					format,
+					blend: Some(wgpu::BlendState::REPLACE),
+					write_mask: wgpu::ColorWrites::ALL,
+				})],
+			}),
+			multiview: None,
+			cache: None,
+		});
+		let pipeline2 = device.create_render_pipeline(&wgpu::RenderPipelineDescriptor {
+			label: None,
+			layout: None,
+			vertex: vertex_state.clone(),
+			primitive: wgpu::PrimitiveState {
+				topology: wgpu::PrimitiveTopology::TriangleStrip,
+				..Default::default()
+			},
+			depth_stencil: None,
+			multisample: wgpu::MultisampleState::default(),
+			fragment: Some(wgpu::FragmentState {
+				module: &shader,
+				entry_point: Some("pass_v"),
+				compilation_options: wgpu::PipelineCompilationOptions::default(),
+				targets: &[Some(wgpu::ColorTargetState {
+					format,
+					blend: Some(wgpu::BlendState {
+						color: wgpu::BlendComponent {
+							src_factor: wgpu::BlendFactor::One,
+							dst_factor: wgpu::BlendFactor::One,
+							operation: wgpu::BlendOperation::Add,
+						},
+						alpha: wgpu::BlendComponent {
+							src_factor: wgpu::BlendFactor::One,
+							dst_factor: wgpu::BlendFactor::One,
+							operation: wgpu::BlendOperation::Add,
+						},
+					}),
+					write_mask: wgpu::ColorWrites::ALL,
+				})],
+			}),
+			multiview: None,
+			cache: None,
+		});
+		Self {
+			view_buf,
+			format,
+			pipeline_h: pipeline1,
+			pipeline_v: pipeline2,
+		}
+	}
+
+	pub fn render(
+		&self,
+		device: &wgpu::Device,
+		encoder: &mut wgpu::CommandEncoder,
+		source: &wgpu::Texture,
+		// target: &wgpu::Texture,
+		params: Params,
+	) {
+		let target = source;
+		let size = target.size();
+		// assert_eq!(source.size(), size);
+		let intermediate = device.create_texture(&wgpu::TextureDescriptor {
+			label: None,
+			size,
+			mip_level_count: 1,
+			sample_count: 1,
+			dimension: wgpu::TextureDimension::D2,
+			format: self.format,
+			usage: wgpu::TextureUsages::TEXTURE_BINDING | wgpu::TextureUsages::RENDER_ATTACHMENT,
+			view_formats: &[],
+		});
+		let intermediate = intermediate.create_view(&wgpu::TextureViewDescriptor::default());
+		let params_buf = device.create_buffer_init(&wgpu::util::BufferInitDescriptor {
+			label: None,
+			contents: bytes_of(&params),
+			usage: wgpu::BufferUsages::UNIFORM,
+		});
+		let source = &source.create_view(&wgpu::TextureViewDescriptor::default());
+		let target = &target.create_view(&wgpu::TextureViewDescriptor::default());
+		{
+			let mut pass = encoder.begin_render_pass(&wgpu::RenderPassDescriptor {
+				label: None,
+				color_attachments: &[Some(wgpu::RenderPassColorAttachment {
+					view: &intermediate,
+					resolve_target: None,
+					ops: wgpu::Operations {
+						load: wgpu::LoadOp::Clear(wgpu::Color::TRANSPARENT),
+						store: wgpu::StoreOp::Store,
+					},
+				})],
+				depth_stencil_attachment: None,
+				occlusion_query_set: None,
+				timestamp_writes: None,
+			});
+			let bindings = device.create_bind_group(&wgpu::BindGroupDescriptor {
+				label: None,
+				layout: &self.pipeline_h.get_bind_group_layout(0),
+				entries: &[
+					wgpu::BindGroupEntry {
+						binding: 1,
+						resource: wgpu::BindingResource::TextureView(source),
+					},
+					wgpu::BindGroupEntry {
+						binding: 2,
+						resource: params_buf.as_entire_binding(),
+					},
+				],
+			});
+			pass.set_pipeline(&self.pipeline_h);
+			pass.set_vertex_buffer(0, self.view_buf.slice(..));
+			pass.set_bind_group(0, &bindings, &[]);
+			pass.draw(0..4, 0..1);
+		}
+		{
+			let mut pass = encoder.begin_render_pass(&wgpu::RenderPassDescriptor {
+				label: None,
+				color_attachments: &[Some(wgpu::RenderPassColorAttachment {
+					view: target,
+					resolve_target: None,
+					ops: wgpu::Operations {
+						// load: wgpu::LoadOp::Load,
+						load: wgpu::LoadOp::Clear(wgpu::Color::TRANSPARENT),
+						store: wgpu::StoreOp::Store,
+					},
+				})],
+				depth_stencil_attachment: None,
+				occlusion_query_set: None,
+				timestamp_writes: None,
+			});
+			let params_buf = device.create_buffer_init(&wgpu::util::BufferInitDescriptor {
+				label: None,
+				contents: bytes_of(&params),
+				usage: wgpu::BufferUsages::UNIFORM,
+			});
+			let bindings = device.create_bind_group(&wgpu::BindGroupDescriptor {
+				label: None,
+				layout: &self.pipeline_v.get_bind_group_layout(0),
+				entries: &[
+					wgpu::BindGroupEntry {
+						binding: 1,
+						resource: wgpu::BindingResource::TextureView(&intermediate),
+					},
+					wgpu::BindGroupEntry {
+						binding: 2,
+						resource: params_buf.as_entire_binding(),
+					},
+				],
+			});
+			pass.set_pipeline(&self.pipeline_v);
+			pass.set_vertex_buffer(0, self.view_buf.slice(..));
+			pass.set_bind_group(0, &bindings, &[]);
+			pass.draw(0..4, 0..1);
+		}
+	}
+}
diff --git a/src/glare.wgsl b/src/glare.wgsl
new file mode 100644
index 0000000..066db36
--- /dev/null
+++ b/src/glare.wgsl
@@ -0,0 +1,92 @@
+struct Params {
+	sigma: f32,
+}
+
+struct Vertex {
+	@location(0) screen: vec2f,
+}
+
+struct Varying {
+	@location(0) tex: vec2f,
+	@builtin(position) v: vec4f,
+}
+
+@group(0) @binding(1) var tex: texture_2d<f32>;
+@group(0) @binding(2) var<uniform> params: Params;
+
+@vertex
+fn on_vertex(in: Vertex) -> Varying {
+	let uv = in.screen;
+	let xy = vec2(2. * uv.x - 1., 1. - 2. * uv.y);
+	return Varying(uv, vec4(xy, 0., 1.));
+}
+
+const CUTOFF: i32 = 200;
+
+@fragment
+fn pass_h(in: Varying) -> @location(0) vec4f {
+	let size = vec2i(textureDimensions(tex));
+	let pix = vec2i(floor(in.tex * vec2f(size)));
+	let a = max(0, pix.x - CUTOFF);
+	let b = min(size.x, pix.x + CUTOFF + 1);
+	var accum = vec4(0.);
+	for (var x = a; x < b; x++) {
+		let w = weight(x - pix.x);
+		let pixel = textureLoad(tex, vec2i(x, pix.y), 0);
+		accum += w * pixel;
+	}
+	return vec4(accum.xyz, params.sigma);
+}
+
+@fragment
+fn pass_v(in: Varying) -> @location(0) vec4f {
+	let size = vec2i(textureDimensions(tex));
+	let pix = vec2i(floor(in.tex * vec2f(size)));
+	let a = max(0, pix.y - CUTOFF);
+	let b = min(size.y, pix.y + CUTOFF + 1);
+	var accum = vec4(0.);
+	for (var y = a; y < b; y++) {
+		let w = weight(y - pix.y);
+		let pixel = textureLoad(tex, vec2i(pix.x, y), 0);
+		accum += w * pixel;
+	}
+	return vec4(accum.xyz, params.sigma);
+}
+
+// fn weight(dist: i32) -> f32 {
+// 	let σ = 50.;
+// 	let d = f32(dist);
+// 	let w = max(0., 1. - abs(d) / σ) / σ;
+// 	return .1 * w;
+// }
+
+fn weight(dist: i32) -> f32 {
+	let σ = params.sigma / 1e5;
+	let d = f32(dist) / σ;
+	return 3. / (1. + d*d);
+}
+
+// fn weight(dist: i32) -> f32 {
+// 	let σ = params.sigma / 1e8;
+// 	let d = f32(dist) / σ;
+// 	return 3. / (1. + abs(d));
+// }
+
+// fn weight(dist: i32) -> f32 {
+// 	let σ = params.sigma / 300.;
+// 	let d = abs(f32(dist) / σ);
+// 	return 1. / (1. + d*d*d) / σ;
+// }
+
+// fn weight(dist: i32) -> f32 {
+// 	let d = f32(abs(dist));
+// 	let c = mix(.01, 1., exp(-d*d));
+// // 	return c * pow(1.2, -d);
+// 	return c / (1. + pow(d, 2.0));
+// // 	return .25 * c / (1. + pow(.1 * d, 4.0));
+// }
+
+// fn weight(dist: i32) -> f32 {
+// 	let d = 2. * f32(dist);
+// 	return exp(-d*d);
+// }
diff --git a/src/lib.rs b/src/lib.rs
index dfaedd1..83f62ae 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -1,4 +1,5 @@
 pub mod anim;
+pub mod glare;
 pub mod perlin;
 pub mod present;
 pub mod scene;
diff --git a/src/present.wgsl b/src/present.wgsl
index 2a7412f..2ea9789 100644
--- a/src/present.wgsl
+++ b/src/present.wgsl
@@ -24,7 +24,7 @@ fn on_vertex(in: Vertex) -> Varying {
 
 @fragment
 fn on_fragment(in: Varying) -> @location(0) vec4f {
-	let pixel = textureSample(tex, smp, in.tex) / params.divisor;
+	let pixel = .1 * textureSample(tex, smp, in.tex) / params.divisor;
 	return vec4(rational_tone_map(pixel.xyz), 1.0);
 }
 
diff --git a/src/scene.rs b/src/scene.rs
index 437cd1b..7c22f20 100644
--- a/src/scene.rs
+++ b/src/scene.rs
@@ -105,7 +105,7 @@ impl Renderer {
 			trace::Params {
 				max_reflections: 3,
 				min_strength: 0.1,
-				sphere_count: N_SPHERES,
+				sphere_count: scene.spheres.len() as u32,
 				seed,
 			},
 			viewport,