diff --git a/Extensions/Sprite.cpp b/Extensions/Sprite.cpp
index f3d7e3e..c82024c 100644
--- a/Extensions/Sprite.cpp
+++ b/Extensions/Sprite.cpp
@@ -868,12 +868,46 @@ void  TFT_eSprite::pushImage(int32_t x, int32_t y, int32_t w, int32_t h, uint16_
   }
   else if (_bpp == 4)
   {
-    // not supported.  The image is unlikely to have the correct colors for the color map.
-    // we could implement a way to push a 4-bit image using the color map?
-    #ifdef TFT_eSPI_DEBUG
-    Serial.println("pushImage(int32_t x, int32_t y, int32_t w, int32_t h, uint16_t *data) not implemented");
-    #endif
-    return;
+    // the image is assumed to be 4 bit, where each byte corresponds to two pixels.
+    // much faster when aligned to a byte boundary, because the alternative is slower, requiring
+    // tedious bit operations.
+
+    const uint8_t *dataBuf = (uint8_t *)data;
+    int sWidth = (_iwidth >> 1);
+    
+    if ((xs & 0x01) == 0 && (xo & 0x01) == 0 && (ws & 0x01) == 0)
+    {
+      if ((ws & 0x01) == 0) // use memcpy for better perf.
+      {
+        xs = (xs >> 1) + ys * sWidth;
+        ws = (ws >> 1);
+        xo = (xo >> 1) + yo * (w>>1);
+        while (hs--)
+        {
+          memcpy(_img4 + xs, dataBuf + xo, ws);
+          xo += (w >> 1);
+          xs += sWidth;
+        }
+      }
+    }
+    else  // not optimized
+    {
+      for (int32_t yp = yo; yp < yo + hs; yp++)
+      {
+        x = xs;
+        for (int32_t xp = xo; xp < xo + ws; xp++)
+        {
+          uint32_t color;
+          if ((xp & 0x01) == 0)
+            color = (dataBuf[((xp+yp*w)>>1)] & 0xF0) >> 4; // even index = bits 7 .. 4
+          else
+            color = dataBuf[((xp-1+yp*w)>>1)] & 0x0F;  // odd index = bits 3 .. 0.
+          drawPixel(x, ys, color);
+          x++; 
+        }
+        ys++;
+      }
+    }
   }
 
   else // 1bpp
diff --git a/Extensions/Sprite.h b/Extensions/Sprite.h
index 74ba6b2..b1307b6 100644
--- a/Extensions/Sprite.h
+++ b/Extensions/Sprite.h
@@ -114,7 +114,7 @@ class TFT_eSprite : public TFT_eSPI {
            // 16bpp = colour, 8bpp = byte, 4bpp = colour index, 1bpp = 1 or 0
   uint16_t readPixelValue(int32_t x, int32_t y);
 
-           // Write an image (colour bitmap) to the sprite.  Not implemented for _bpp == 4.
+           // Write an image (colour bitmap) to the sprite.
   void     pushImage(int32_t x0, int32_t y0, int32_t w, int32_t h, uint16_t *data);
   void     pushImage(int32_t x0, int32_t y0, int32_t w, int32_t h, const uint16_t *data);
 
diff --git a/examples/Sprite/Sprite_image_4bit/Sprite_image_4bit.ino b/examples/Sprite/Sprite_image_4bit/Sprite_image_4bit.ino
index e004df4..bdf5390 100644
--- a/examples/Sprite/Sprite_image_4bit/Sprite_image_4bit.ino
+++ b/examples/Sprite/Sprite_image_4bit/Sprite_image_4bit.ino
@@ -39,10 +39,6 @@ TFT_eSPI    tft = TFT_eSPI();         // Declare object "tft"
 
 TFT_eSprite spr = TFT_eSprite(&tft);  // Declare Sprite object "spr" with pointer to "tft" object
 
-const int freq = 5000;
-int screenBrightnessChannel = 0;
-int resolution = 8;
-
 byte red = 31; // Red is the top 5 bits of a 16 bit colour value
 byte green = 0;// Green is the middle 6 bits
 byte blue = 0; // Blue is the bottom 5 bits
@@ -55,11 +51,6 @@ uint16_t cmap[16];
 
 void setup()
 {
-  pinMode(21, OUTPUT);
-  ledcSetup(screenBrightnessChannel, freq, resolution);
-  ledcAttachPin(21, screenBrightnessChannel);
-  ledcWrite(screenBrightnessChannel, 127);
-
   Serial.begin(9600);
   Serial.println();
 
@@ -94,13 +85,12 @@ void loop(void)
     cmap[i] = cmap[i + 1];
   }
   if (incr == 2) {
-    (void)rainbow();  // skip alternate steps
+    (void)rainbow();  // skip alternate steps to go faster
   }
   cmap[15] = rainbow();
   rloop += incr;
   if (rloop > 0xc0) {
     incr = incr == 2 ? 1 : 2;
-    Serial.printf("incr %d, rloop %d\r\n", incr, rloop);
     rloop = 0;
     
   }