diff --git a/README.md b/README.md
index 22bfc56..792f34b 100644
--- a/README.md
+++ b/README.md
@@ -1,4 +1,4 @@
-# PCB to Stencil Converter
+# Gerber Solder Paste Layer to Solder Stencil Converter
 
 A Go tool to convert Gerber files (specifically solder paste layers) into 3D printable STL stencils.
 
@@ -20,8 +20,10 @@ go run main.go gerber.go [options] <path_to_gerber_file>
 
 ### Options
 
-- `--height, -h`: Stencil height in mm (default: 0.2mm).
-- `--keep-png, --kp`: Save the intermediate PNG image used for mesh generation (useful for debugging).
+- `--height`: Stencil height in mm (default: 0.2mm).
+- `--wall-height`: Wall height mm (default: 2.0mm).
+- `--wall-thickness`: Wall thickness in mm (default: 1mm).
+- `--keep-png`: Save the intermediate PNG image used for mesh generation (useful for debugging).
 
 ### Example
 
diff --git a/bin/pcb-to-stencil b/bin/pcb-to-stencil
index 9a02db3..c4fc467 100755
Binary files a/bin/pcb-to-stencil and b/bin/pcb-to-stencil differ
diff --git a/gerber.go b/gerber.go
index 42cdd0f..e1888d5 100644
--- a/gerber.go
+++ b/gerber.go
@@ -219,9 +219,11 @@ func (gf *GerberFile) parseCoordinate(valStr string, fmtSpec struct{ Integer, De
 	return val / divisor
 }
 
-// Render generates an image from the parsed Gerber commands
-func (gf *GerberFile) Render(dpi float64) image.Image {
-	// 1. Calculate Bounds
+type Bounds struct {
+	MinX, MinY, MaxX, MaxY float64
+}
+
+func (gf *GerberFile) CalculateBounds() Bounds {
 	minX, minY := 1e9, 1e9
 	maxX, maxY := -1e9, -1e9
 
@@ -271,8 +273,20 @@ func (gf *GerberFile) Render(dpi float64) image.Image {
 	maxX += padding
 	maxY += padding
 
-	widthMM := maxX - minX
-	heightMM := maxY - minY
+	return Bounds{MinX: minX, MinY: minY, MaxX: maxX, MaxY: maxY}
+}
+
+// Render generates an image from the parsed Gerber commands
+func (gf *GerberFile) Render(dpi float64, bounds *Bounds) image.Image {
+	var b Bounds
+	if bounds != nil {
+		b = *bounds
+	} else {
+		b = gf.CalculateBounds()
+	}
+
+	widthMM := b.MaxX - b.MinX
+	heightMM := b.MaxY - b.MinY
 
 	var scale float64
 	if gf.State.Units == "IN" {
@@ -294,12 +308,12 @@ func (gf *GerberFile) Render(dpi float64) image.Image {
 
 	// Helper to convert mm to pixels
 	toPix := func(x, y float64) (int, int) {
-		px := int((x - minX) * scale)
-		py := int((heightMM - (y - minY)) * scale) // Flip Y for image coords
+		px := int((x - b.MinX) * scale)
+		py := int((heightMM - (y - b.MinY)) * scale) // Flip Y for image coords
 		return px, py
 	}
 
-	curX, curY = 0.0, 0.0
+	curX, curY := 0.0, 0.0
 	curDCode := 0
 
 	for _, cmd := range gf.Commands {
@@ -343,7 +357,7 @@ func (gf *GerberFile) drawAperture(img *image.RGBA, x, y int, ap Aperture, scale
 		// Modifiers[0] is diameter
 		if len(ap.Modifiers) > 0 {
 			radius := int((ap.Modifiers[0] * scale) / 2)
-			drawCircle(img, x, y, radius, c)
+			drawCircle(img, x, y, radius)
 		}
 		return
 	case ApertureRect: // R
@@ -383,7 +397,7 @@ func (gf *GerberFile) drawAperture(img *image.RGBA, x, y int, ap Aperture, scale
 					py := int(cy * scale)
 
 					radius := int((dia * scale) / 2)
-					drawCircle(img, x+px, y-py, radius, c)
+					drawCircle(img, x+px, y-py, radius)
 				}
 			case 21: // Center Line (Rect)
 				// Mods: Exposure, Width, Height, CenterX, CenterY, Rotation
@@ -421,7 +435,7 @@ func (gf *GerberFile) drawAperture(img *image.RGBA, x, y int, ap Aperture, scale
 	}
 }
 
-func drawCircle(img *image.RGBA, x0, y0, r int, c image.Image) {
+func drawCircle(img *image.RGBA, x0, y0, r int) {
 	// Simple Bresenham or scanline
 	for y := -r; y <= r; y++ {
 		for x := -r; x <= r; x++ {
diff --git a/main.go b/main.go
index 0a9e734..7d3fb02 100644
--- a/main.go
+++ b/main.go
@@ -1,23 +1,25 @@
 package main
 
 import (
+	"encoding/binary"
 	"flag"
 	"fmt"
 	"image"
 	"image/png"
 	"log"
+	"math"
 	"os"
 	"path/filepath"
 	"strings"
 )
 
 // --- Configuration ---
-const (
-	DPI       = 1000.0 // Higher DPI = smoother curves
-	PixelToMM = 25.4 / DPI
-)
+var DPI float64 = 1000.0 // Higher DPI = smoother curves
+var PixelToMM float64 = 25.4 / DPI
 
 var StencilHeight float64 = 0.2 // mm, default
+var WallHeight float64 = 2.0    // mm, default
+var WallThickness float64 = 1.0 // mm, default
 var KeepPNG bool
 
 // --- STL Helpers ---
@@ -33,18 +35,58 @@ func WriteSTL(filename string, triangles [][3]Point) error {
 	}
 	defer f.Close()
 
-	// Writing Binary STL is harder, ASCII is fine for this size
-	f.WriteString("solid stencil\n")
-	for _, t := range triangles {
-		f.WriteString("facet normal 0 0 0\n")
-		f.WriteString("  outer loop\n")
-		for _, p := range t {
-			f.WriteString(fmt.Sprintf("    vertex %f %f %f\n", p.X, p.Y, p.Z))
-		}
-		f.WriteString("  endloop\n")
-		f.WriteString("endfacet\n")
+	// Write Binary STL Header (80 bytes)
+	header := make([]byte, 80)
+	copy(header, "Generated by pcb-to-stencil")
+	if _, err := f.Write(header); err != nil {
+		return err
+	}
+
+	// Write Number of Triangles (4 bytes uint32)
+	count := uint32(len(triangles))
+	if err := binary.Write(f, binary.LittleEndian, count); err != nil {
+		return err
+	}
+
+	// Write Triangles
+	// Each triangle is 50 bytes:
+	// Normal (3 floats = 12 bytes)
+	// Vertex 1 (3 floats = 12 bytes)
+	// Vertex 2 (3 floats = 12 bytes)
+	// Vertex 3 (3 floats = 12 bytes)
+	// Attribute byte count (2 bytes uint16)
+
+	// Buffer for a single triangle to minimize syscalls
+	buf := make([]byte, 50)
+
+	for _, t := range triangles {
+		// Normal (0,0,0)
+		binary.LittleEndian.PutUint32(buf[0:4], math.Float32bits(0))
+		binary.LittleEndian.PutUint32(buf[4:8], math.Float32bits(0))
+		binary.LittleEndian.PutUint32(buf[8:12], math.Float32bits(0))
+
+		// Vertex 1
+		binary.LittleEndian.PutUint32(buf[12:16], math.Float32bits(float32(t[0].X)))
+		binary.LittleEndian.PutUint32(buf[16:20], math.Float32bits(float32(t[0].Y)))
+		binary.LittleEndian.PutUint32(buf[20:24], math.Float32bits(float32(t[0].Z)))
+
+		// Vertex 2
+		binary.LittleEndian.PutUint32(buf[24:28], math.Float32bits(float32(t[1].X)))
+		binary.LittleEndian.PutUint32(buf[28:32], math.Float32bits(float32(t[1].Y)))
+		binary.LittleEndian.PutUint32(buf[32:36], math.Float32bits(float32(t[1].Z)))
+
+		// Vertex 3
+		binary.LittleEndian.PutUint32(buf[36:40], math.Float32bits(float32(t[2].X)))
+		binary.LittleEndian.PutUint32(buf[40:44], math.Float32bits(float32(t[2].Y)))
+		binary.LittleEndian.PutUint32(buf[44:48], math.Float32bits(float32(t[2].Z)))
+
+		// Attribute byte count (0)
+		binary.LittleEndian.PutUint16(buf[48:50], 0)
+
+		if _, err := f.Write(buf); err != nil {
+			return err
+		}
 	}
-	f.WriteString("endsolid stencil\n")
 	return nil
 }
 
@@ -77,27 +119,283 @@ func AddBox(triangles *[][3]Point, x, y, w, h, zHeight float64) {
 
 // --- Meshing Logic (Optimized) ---
 
-func GenerateMeshFromImage(img image.Image) [][3]Point {
+// ComputeWallMask generates a mask for the wall based on the outline image.
+// It identifies the board area (inside the outline) and creates a wall of
+// specified thickness around it.
+func ComputeWallMask(img image.Image, thicknessMM float64) ([]bool, []bool) {
 	bounds := img.Bounds()
+	w := bounds.Max.X
+	h := bounds.Max.Y
+	size := w * h
+
+	// Helper for neighbors
+	dx := []int{0, 0, 1, -1}
+	dy := []int{1, -1, 0, 0}
+
+	// 1. Identify Outline Pixels (White)
+	isOutline := make([]bool, size)
+	outlineQueue := []int{}
+	for i := 0; i < size; i++ {
+		cx := i % w
+		cy := i / w
+		c := img.At(cx, cy)
+		r, _, _, _ := c.RGBA()
+		if r > 10000 { // White-ish
+			isOutline[i] = true
+			outlineQueue = append(outlineQueue, i)
+		}
+	}
+
+	// 2. Dilate Outline to close gaps
+	// We dilate by a small amount (e.g. 0.5mm) to ensure the outline is closed.
+	gapClosingMM := 0.5
+	gapClosingPixels := int(gapClosingMM / PixelToMM)
+	if gapClosingPixels < 1 {
+		gapClosingPixels = 1
+	}
+
+	dist := make([]int, size)
+	for i := 0; i < size; i++ {
+		if isOutline[i] {
+			dist[i] = 0
+		} else {
+			dist[i] = -1
+		}
+	}
+
+	// BFS for Dilation
+	dilatedOutline := make([]bool, size)
+	copy(dilatedOutline, isOutline)
+
+	// Use a separate queue for dilation to avoid modifying the original outlineQueue if we needed it
+	dQueue := make([]int, len(outlineQueue))
+	copy(dQueue, outlineQueue)
+
+	for len(dQueue) > 0 {
+		idx := dQueue[0]
+		dQueue = dQueue[1:]
+
+		d := dist[idx]
+		if d >= gapClosingPixels {
+			continue
+		}
+
+		cx := idx % w
+		cy := idx / w
+
+		for i := 0; i < 4; i++ {
+			nx, ny := cx+dx[i], cy+dy[i]
+			if nx >= 0 && nx < w && ny >= 0 && ny < h {
+				nIdx := ny*w + nx
+				if dist[nIdx] == -1 {
+					dist[nIdx] = d + 1
+					dilatedOutline[nIdx] = true
+					dQueue = append(dQueue, nIdx)
+				}
+			}
+		}
+	}
+
+	// 3. Flood Fill "Outside" using Dilated Outline as barrier
+	isOutside := make([]bool, size)
+	// Start from (0,0) - assumed to be outside due to padding
+	if !dilatedOutline[0] {
+		isOutside[0] = true
+		fQueue := []int{0}
+
+		for len(fQueue) > 0 {
+			idx := fQueue[0]
+			fQueue = fQueue[1:]
+
+			cx := idx % w
+			cy := idx / w
+
+			for i := 0; i < 4; i++ {
+				nx, ny := cx+dx[i], cy+dy[i]
+				if nx >= 0 && nx < w && ny >= 0 && ny < h {
+					nIdx := ny*w + nx
+					if !isOutside[nIdx] && !dilatedOutline[nIdx] {
+						isOutside[nIdx] = true
+						fQueue = append(fQueue, nIdx)
+					}
+				}
+			}
+		}
+	}
+
+	// 4. Restore Board Shape (Erode "Outside" back to original boundary)
+	// We dilated the outline, so "Outside" stopped 'gapClosingPixels' away from the real board edge.
+	// We need to expand "Outside" inwards by 'gapClosingPixels' to touch the real board edge.
+	// Then "Board" = !Outside.
+
+	// Reset dist for Outside expansion
+	for i := 0; i < size; i++ {
+		if isOutside[i] {
+			dist[i] = 0
+		} else {
+			dist[i] = -1
+		}
+	}
+
+	oQueue := []int{}
+	for i := 0; i < size; i++ {
+		if isOutside[i] {
+			oQueue = append(oQueue, i)
+		}
+	}
+
+	isOutsideExpanded := make([]bool, size)
+	copy(isOutsideExpanded, isOutside)
+
+	for len(oQueue) > 0 {
+		idx := oQueue[0]
+		oQueue = oQueue[1:]
+
+		d := dist[idx]
+		if d >= gapClosingPixels {
+			continue
+		}
+
+		cx := idx % w
+		cy := idx / w
+
+		for i := 0; i < 4; i++ {
+			nx, ny := cx+dx[i], cy+dy[i]
+			if nx >= 0 && nx < w && ny >= 0 && ny < h {
+				nIdx := ny*w + nx
+				if dist[nIdx] == -1 {
+					dist[nIdx] = d + 1
+					isOutsideExpanded[nIdx] = true
+					oQueue = append(oQueue, nIdx)
+				}
+			}
+		}
+	}
+
+	// 5. Define Board
+	isBoard := make([]bool, size)
+	for i := 0; i < size; i++ {
+		isBoard[i] = !isOutsideExpanded[i]
+	}
+
+	// 6. Generate Wall
+	// Wall is generated by expanding Board outwards.
+	// We want the wall to be strictly OUTSIDE the board (or centered on outline? User said "starts at outline").
+	// If we expand Board, we get pixels outside.
+
+	thicknessPixels := int(thicknessMM / PixelToMM)
+	if thicknessPixels < 1 {
+		thicknessPixels = 1
+	}
+
+	// Reset dist for Wall generation
+	for i := 0; i < size; i++ {
+		if isBoard[i] {
+			dist[i] = 0
+		} else {
+			dist[i] = -1
+		}
+	}
+
+	wQueue := []int{}
+	for i := 0; i < size; i++ {
+		if isBoard[i] {
+			wQueue = append(wQueue, i)
+		}
+	}
+
+	isWall := make([]bool, size)
+
+	for len(wQueue) > 0 {
+		idx := wQueue[0]
+		wQueue = wQueue[1:]
+
+		d := dist[idx]
+		if d >= thicknessPixels {
+			continue
+		}
+
+		cx := idx % w
+		cy := idx / w
+
+		for i := 0; i < 4; i++ {
+			nx, ny := cx+dx[i], cy+dy[i]
+			if nx >= 0 && nx < w && ny >= 0 && ny < h {
+				nIdx := ny*w + nx
+				if dist[nIdx] == -1 {
+					dist[nIdx] = d + 1
+					isWall[nIdx] = true
+					wQueue = append(wQueue, nIdx)
+				}
+			}
+		}
+	}
+
+	return isWall, isBoard
+}
+
+func GenerateMeshFromImages(stencilImg, outlineImg image.Image) [][3]Point {
+	bounds := stencilImg.Bounds()
 	width := bounds.Max.X
 	height := bounds.Max.Y
 	var triangles [][3]Point
 
+	var wallMask []bool
+	var boardMask []bool
+	if outlineImg != nil {
+		fmt.Println("Computing wall mask...")
+		wallMask, boardMask = ComputeWallMask(outlineImg, WallThickness)
+	}
+
 	// Optimization: Run-Length Encoding
 	for y := 0; y < height; y++ {
 		var startX = -1
+		var currentHeight = 0.0
 
 		for x := 0; x < width; x++ {
-			c := img.At(x, y)
-			r, g, b, _ := c.RGBA()
+			// Check stencil (black = solid)
+			sc := stencilImg.At(x, y)
+			sr, sg, sb, _ := sc.RGBA()
+			isStencilSolid := sr < 10000 && sg < 10000 && sb < 10000
 
-			// Check for BLACK pixels (The Plastic Stencil Body)
-			// Adjust threshold if gerbv produces slightly gray blacks
-			isSolid := r < 10000 && g < 10000 && b < 10000
+			// Check wall
+			isWall := false
+			isInsideBoard := true
+			if wallMask != nil {
+				idx := y*width + x
+				isWall = wallMask[idx]
+				if boardMask != nil {
+					isInsideBoard = boardMask[idx]
+				}
+			}
 
-			if isSolid {
+			// Determine height at this pixel
+			h := 0.0
+			if isWall {
+				h = WallHeight
+			} else if isStencilSolid {
+				if isInsideBoard {
+					h = StencilHeight
+				}
+			}
+
+			if h > 0 {
 				if startX == -1 {
 					startX = x
+					currentHeight = h
+				} else if h != currentHeight {
+					// Height changed, end current strip and start new one
+					stripLen := x - startX
+					AddBox(
+						&triangles,
+						float64(startX)*PixelToMM,
+						float64(y)*PixelToMM,
+						float64(stripLen)*PixelToMM,
+						PixelToMM,
+						currentHeight,
+					)
+					startX = x
+					currentHeight = h
 				}
 			} else {
 				if startX != -1 {
@@ -109,9 +407,10 @@ func GenerateMeshFromImage(img image.Image) [][3]Point {
 						float64(y)*PixelToMM,
 						float64(stripLen)*PixelToMM,
 						PixelToMM,
-						StencilHeight,
+						currentHeight,
 					)
 					startX = -1
+					currentHeight = 0.0
 				}
 			}
 		}
@@ -123,7 +422,7 @@ func GenerateMeshFromImage(img image.Image) [][3]Point {
 				float64(y)*PixelToMM,
 				float64(stripLen)*PixelToMM,
 				PixelToMM,
-				StencilHeight,
+				currentHeight,
 			)
 		}
 	}
@@ -134,33 +433,83 @@ func GenerateMeshFromImage(img image.Image) [][3]Point {
 
 func main() {
 	flag.Float64Var(&StencilHeight, "height", 0.2, "Stencil height in mm")
-	flag.Float64Var(&StencilHeight, "h", 0.2, "Stencil height in mm (short)")
+	flag.Float64Var(&WallHeight, "wall-height", 2.0, "Wall height in mm")
+	flag.Float64Var(&WallThickness, "wall-thickness", 1, "Wall thickness in mm")
+	flag.Float64Var(&DPI, "dpi", 1000.0, "DPI for rendering (lower = smaller file, rougher curves)")
 	flag.BoolVar(&KeepPNG, "keep-png", false, "Save intermediate PNG file")
-	flag.BoolVar(&KeepPNG, "kp", false, "Save intermediate PNG file (short)")
 	flag.Parse()
 
+	// Update PixelToMM based on DPI flag
+	PixelToMM = 25.4 / DPI
+
 	args := flag.Args()
 	if len(args) < 1 {
-		fmt.Println("Usage: go run main.go [options] <path_to_gerber_file>")
+		fmt.Println("Usage: go run main.go [options] <path_to_gerber_file> [path_to_outline_gerber_file]")
 		fmt.Println("Options:")
 		flag.PrintDefaults()
-		fmt.Println("Example: go run main.go -height=0.3 MyPCB.GTP")
+		fmt.Println("Example: go run main.go -height=0.3 MyPCB.GTP MyPCB.GKO")
 		os.Exit(1)
 	}
 
 	gerberPath := args[0]
+	var outlinePath string
+	if len(args) > 1 {
+		outlinePath = args[1]
+	}
+
 	outputPath := strings.TrimSuffix(gerberPath, filepath.Ext(gerberPath)) + ".stl"
 
-	// 1. Parse Gerber
+	// 1. Parse Gerber(s)
 	fmt.Printf("Parsing %s...\n", gerberPath)
 	gf, err := ParseGerber(gerberPath)
 	if err != nil {
 		log.Fatalf("Error parsing gerber: %v", err)
 	}
 
-	// 2. Render to Image
+	var outlineGf *GerberFile
+	if outlinePath != "" {
+		fmt.Printf("Parsing outline %s...\n", outlinePath)
+		outlineGf, err = ParseGerber(outlinePath)
+		if err != nil {
+			log.Fatalf("Error parsing outline gerber: %v", err)
+		}
+	}
+
+	// 2. Calculate Union Bounds
+	bounds := gf.CalculateBounds()
+	if outlineGf != nil {
+		outlineBounds := outlineGf.CalculateBounds()
+		if outlineBounds.MinX < bounds.MinX {
+			bounds.MinX = outlineBounds.MinX
+		}
+		if outlineBounds.MinY < bounds.MinY {
+			bounds.MinY = outlineBounds.MinY
+		}
+		if outlineBounds.MaxX > bounds.MaxX {
+			bounds.MaxX = outlineBounds.MaxX
+		}
+		if outlineBounds.MaxY > bounds.MaxY {
+			bounds.MaxY = outlineBounds.MaxY
+		}
+	}
+
+	// Expand bounds to accommodate wall thickness and prevent clipping
+	// We add WallThickness + extra margin to all sides
+	margin := WallThickness + 5.0 // mm
+	bounds.MinX -= margin
+	bounds.MinY -= margin
+	bounds.MaxX += margin
+	bounds.MaxY += margin
+
+	// 3. Render to Image(s)
 	fmt.Println("Rendering to internal image...")
-	img := gf.Render(DPI)
+	img := gf.Render(DPI, &bounds)
+
+	var outlineImg image.Image
+	if outlineGf != nil {
+		fmt.Println("Rendering outline to internal image...")
+		outlineImg = outlineGf.Render(DPI, &bounds)
+	}
 
 	if KeepPNG {
 		pngPath := strings.TrimSuffix(gerberPath, filepath.Ext(gerberPath)) + ".png"
@@ -174,13 +523,27 @@ func main() {
 			}
 			f.Close()
 		}
+
+		if outlineImg != nil {
+			outlinePngPath := strings.TrimSuffix(gerberPath, filepath.Ext(gerberPath)) + "_outline.png"
+			fmt.Printf("Saving intermediate Outline PNG to %s...\n", outlinePngPath)
+			f, err := os.Create(outlinePngPath)
+			if err != nil {
+				log.Printf("Warning: Could not create Outline PNG file: %v", err)
+			} else {
+				if err := png.Encode(f, outlineImg); err != nil {
+					log.Printf("Warning: Could not encode Outline PNG: %v", err)
+				}
+				f.Close()
+			}
+		}
 	}
 
-	// 3. Generate Mesh
+	// 4. Generate Mesh
 	fmt.Println("Generating mesh (this may take 10-20 seconds for large boards)...")
-	triangles := GenerateMeshFromImage(img)
+	triangles := GenerateMeshFromImages(img, outlineImg)
 
-	// 4. Save STL
+	// 5. Save STL
 	fmt.Printf("Saving to %s (%d triangles)...\n", outputPath, len(triangles))
 	err = WriteSTL(outputPath, triangles)
 	if err != nil {