diff --git a/wled00/colors.cpp b/wled00/colors.cpp
index f741e7a6..f1bbea0a 100644
--- a/wled00/colors.cpp
+++ b/wled00/colors.cpp
@@ -102,6 +102,40 @@ uint32_t WLED_O2_ATTR IRAM_ATTR_YN __attribute__((hot)) color_fade(uint32_t c1,
   uint32_t wg = (((c1 >> 8) & TWO_CHANNEL_MASK) * amount) & ~TWO_CHANNEL_MASK; // scale white and green
   return (rb | wg) + addRemains;
 }
+
+//
+// overwrite FastLed colorFromPalette with an optimized version created by dedehai (https://github.com/Aircoookie/WLED/pull/4138)
+// 
+// 1:1 replacement of fastled function optimized for ESP, slightly faster, more accurate and uses less flash (~ -200bytes)
+CRGB IRAM_ATTR_YN __attribute__((hot)) ColorFromPaletteWLED(const CRGBPalette16& pal, unsigned index, uint8_t brightness, TBlendType blendType)
+{
+  if (blendType == LINEARBLEND_NOWRAP) {
+    index = (index*240) >> 8; // Blend range is affected by lo4 blend of values, remap to avoid wrapping
+  }
+  unsigned hi4 = byte(index) >> 4;
+  const CRGB* entry = (CRGB*)((uint8_t*)(&(pal[0])) + (hi4 * sizeof(CRGB)));
+  unsigned red1   = entry->r;
+  unsigned green1 = entry->g;
+  unsigned blue1  = entry->b;
+  if (blendType != NOBLEND) {
+    if (hi4 == 15) entry = &(pal[0]);
+    else ++entry;
+    unsigned f2 = ((index & 0x0F) << 4) + 1; // +1 so we scale by 256 as a max value, then result can just be shifted by 8
+    unsigned f1 = (257 - f2); // f2 is 1 minimum, so this is 256 max
+    red1   = (red1 * f1 + (unsigned)entry->r * f2) >> 8;
+    green1 = (green1 * f1 + (unsigned)entry->g * f2) >> 8;
+    blue1  = (blue1 * f1 + (unsigned)entry->b * f2) >> 8;
+  }
+  if (brightness < 255) { // note: zero checking could be done to return black but that is hardly ever used so it is omitted
+    uint32_t scale = brightness + 1; // adjust for rounding (bitshift)
+    red1   = (red1 * scale) >> 8;
+    green1 = (green1 * scale) >> 8;
+    blue1  = (blue1 * scale) >> 8;
+  }
+  return RGBW32(red1,green1,blue1,0);
+  //return CRGB(red1,green1,blue1);
+}
+
 #endif
 
 void setRandomColor(byte* rgb)
diff --git a/wled00/fcn_declare.h b/wled00/fcn_declare.h
index 188c10ee..c5e95604 100644
--- a/wled00/fcn_declare.h
+++ b/wled00/fcn_declare.h
@@ -51,9 +51,16 @@ bool getJsonValue(const JsonVariant& element, DestType& destination, const Defau
 
 //colors.cpp
 #if !defined(ARDUINO_ARCH_ESP32) || !defined(WLEDMM_FASTPATH) || defined(WLEDMM_SAVE_FLASH)  // WLEDMM: color utils moved into colorTools.hpp, so the compiler may inline these functions (faster)
+#if !defined(FASTLED_VERSION) // pull in FastLED if we don't have it yet (we need the CRGB type)
+  #define FASTLED_INTERNAL
+  #include <FastLED.h>
+#endif
 uint32_t __attribute__((const)) color_blend(uint32_t,uint32_t,uint_fast16_t,bool b16=false);  // WLEDMM: added attribute const
 uint32_t __attribute__((const)) color_add(uint32_t,uint32_t, bool fast=false);                // WLEDMM: added attribute const
 uint32_t __attribute__((const)) color_fade(uint32_t c1, uint8_t amount, bool video=false);
+#undef ColorFromPalette // overwrite any existing override
+CRGB __attribute__((hot,const)) ColorFromPaletteWLED(const CRGBPalette16& pal, unsigned index, uint8_t brightness=255, TBlendType blendType=LINEARBLEND);
+#define ColorFromPalette ColorFromPaletteWLED // override fastled function
 #else
 #include "colorTools.hpp"
 #endif