diff --git a/Memola/Canvas/Geometries/Stroke/Stroke.swift b/Memola/Canvas/Geometries/Stroke/Stroke.swift
index 32081b2..8018da1 100644
--- a/Memola/Canvas/Geometries/Stroke/Stroke.swift
+++ b/Memola/Canvas/Geometries/Stroke/Stroke.swift
@@ -56,8 +56,9 @@ final class Stroke: @unchecked Sendable {
 
     let movingAverage = MovingAverage(windowSize: 3)
 
-    var vertexBuffer: MTLBuffer?
     var texture: MTLTexture?
+    var indexBuffer: MTLBuffer?
+    var vertexBuffer: MTLBuffer?
 
     var isEmpty: Bool {
         quads.isEmpty
@@ -160,12 +161,19 @@ extension Stroke: Drawable {
     }
 
     func draw(device: MTLDevice, renderEncoder: MTLRenderCommandEncoder) {
-        guard !isEmpty else { return }
+        guard !isEmpty, let indexBuffer else { return }
         prepare(device: device)
         renderEncoder.setFragmentTexture(texture, index: 0)
         renderEncoder.setVertexBuffer(vertexBuffer, offset: 0, index: 0)
-        renderEncoder.drawPrimitives(type: .triangle, vertexStart: 0, vertexCount: quads.endIndex * 6)
-        vertexBuffer = nil
+        renderEncoder.drawIndexedPrimitives(
+            type: .triangle,
+            indexCount: quads.endIndex * 6,
+            indexType: .uint32,
+            indexBuffer: indexBuffer,
+            indexBufferOffset: 0
+        )
+        self.vertexBuffer = nil
+        self.indexBuffer = nil
     }
 }
 
diff --git a/Memola/Canvas/RenderPasses/EraserRenderPass.swift b/Memola/Canvas/RenderPasses/EraserRenderPass.swift
index 46bbe54..32f115d 100644
--- a/Memola/Canvas/RenderPasses/EraserRenderPass.swift
+++ b/Memola/Canvas/RenderPasses/EraserRenderPass.swift
@@ -57,12 +57,15 @@ class EraserRenderPass: RenderPass {
         let quadCount = stroke.quads.endIndex
         var quads = stroke.quads
         let quadBuffer = renderer.device.makeBuffer(bytes: &quads, length: MemoryLayout<Quad>.stride * quadCount, options: [])
-        let vertexBuffer = renderer.device.makeBuffer(length: MemoryLayout<QuadVertex>.stride * quadCount * 6, options: [])
+        let indexBuffer = renderer.device.makeBuffer(length: MemoryLayout<UInt32>.stride * quadCount * 6, options: [])
+        let vertexBuffer = renderer.device.makeBuffer(length: MemoryLayout<QuadVertex>.stride * quadCount * 4, options: [])
 
         computeEncoder.setComputePipelineState(quadPipelineState)
         computeEncoder.setBuffer(quadBuffer, offset: 0, index: 0)
-        computeEncoder.setBuffer(vertexBuffer, offset: 0, index: 1)
+        computeEncoder.setBuffer(indexBuffer, offset: 0, index: 1)
+        computeEncoder.setBuffer(vertexBuffer, offset: 0, index: 2)
 
+        stroke.indexBuffer = indexBuffer
         stroke.vertexBuffer = vertexBuffer
 
         let threadsPerGroup = MTLSize(width: 1, height: 1, depth: 1)
diff --git a/Memola/Canvas/RenderPasses/StrokeRenderPass.swift b/Memola/Canvas/RenderPasses/StrokeRenderPass.swift
index c268735..89945af 100644
--- a/Memola/Canvas/RenderPasses/StrokeRenderPass.swift
+++ b/Memola/Canvas/RenderPasses/StrokeRenderPass.swift
@@ -70,12 +70,15 @@ class StrokeRenderPass: RenderPass {
         let quadCount = stroke.quads.endIndex
         var quads = stroke.quads
         let quadBuffer = renderer.device.makeBuffer(bytes: &quads, length: MemoryLayout<Quad>.stride * quadCount, options: [])
-        let vertexBuffer = renderer.device.makeBuffer(length: MemoryLayout<QuadVertex>.stride * quadCount * 6, options: [])
+        let indexBuffer = renderer.device.makeBuffer(length: MemoryLayout<UInt32>.stride * quadCount * 6, options: [])
+        let vertexBuffer = renderer.device.makeBuffer(length: MemoryLayout<QuadVertex>.stride * quadCount * 4, options: [])
 
         computeEncoder.setComputePipelineState(quadPipelineState)
         computeEncoder.setBuffer(quadBuffer, offset: 0, index: 0)
-        computeEncoder.setBuffer(vertexBuffer, offset: 0, index: 1)
+        computeEncoder.setBuffer(indexBuffer, offset: 0, index: 1)
+        computeEncoder.setBuffer(vertexBuffer, offset: 0, index: 2)
 
+        stroke.indexBuffer = indexBuffer
         stroke.vertexBuffer = vertexBuffer
 
         let threadsPerGroup = MTLSize(width: 1, height: 1, depth: 1)
diff --git a/Memola/Canvas/Shaders/Quad.metal b/Memola/Canvas/Shaders/Quad.metal
index 8bd9cc8..898c3d0 100644
--- a/Memola/Canvas/Shaders/Quad.metal
+++ b/Memola/Canvas/Shaders/Quad.metal
@@ -39,16 +39,23 @@ Vertex createVertex(Quad quad, float2 factor, float2 textCoord) {
 
 kernel void generate_stroke_vertices(
     device Quad *quads [[buffer(0)]],
-    device Vertex *vertices [[buffer(1)]],
+    device uint *indices [[buffer(1)]],
+    device Vertex *vertices [[buffer(2)]],
     uint gid [[thread_position_in_grid]]
 ) {
-    uint index = gid * 6;
     Quad quad = quads[gid];
+    uint index = gid * 6;
+    uint vertexIndex = gid * 4;
     float halfSize = quad.size * 0.5;
-    vertices[index] = createVertex(quad, float2(-halfSize, -halfSize), float2(0, 0));
-    vertices[index + 1] = createVertex(quad, float2(halfSize, -halfSize), float2(1, 0));
-    vertices[index + 2] = createVertex(quad, float2(-halfSize, halfSize), float2(0, 1));
-    vertices[index + 3] = createVertex(quad, float2(halfSize, -halfSize), float2(1, 0));
-    vertices[index + 4] = createVertex(quad, float2(-halfSize, halfSize), float2(0, 1));
-    vertices[index + 5] = createVertex(quad, float2(halfSize, halfSize), float2(1, 1));
+    vertices[vertexIndex] = createVertex(quad, float2(-halfSize, -halfSize), float2(0, 0));
+    vertices[vertexIndex + 1] = createVertex(quad, float2(halfSize, -halfSize), float2(1, 0));
+    vertices[vertexIndex + 2] = createVertex(quad, float2(-halfSize, halfSize), float2(0, 1));
+    vertices[vertexIndex + 3] = createVertex(quad, float2(halfSize, halfSize), float2(1, 1));
+
+    indices[index] = vertexIndex;
+    indices[index + 1] = vertexIndex + 1;
+    indices[index + 2] = vertexIndex + 2;
+    indices[index + 3] = vertexIndex + 1;
+    indices[index + 4] = vertexIndex + 2;
+    indices[index + 5] = vertexIndex + 3;
 }