diff --git a/src/d_main.cpp b/src/d_main.cpp
index ae5400116..dc947187c 100644
--- a/src/d_main.cpp
+++ b/src/d_main.cpp
@@ -1506,6 +1506,8 @@ void D_SRB2Main(void)
 
 #endif //ifndef DEVELOP
 
+	R_InitPaletteRemap();
+
 	// now do it again for the SOC colors in main.pk3!
 	M_InitPlayerSetupColors();
 
diff --git a/src/hardware/hw_cache.c b/src/hardware/hw_cache.c
index c8ae308e4..79fb6f4cb 100644
--- a/src/hardware/hw_cache.c
+++ b/src/hardware/hw_cache.c
@@ -522,12 +522,18 @@ static void HWR_GenerateTexture(GLMapTexture_t *grtex, INT32 texnum, boolean noe
 #endif
 #ifdef WALLFLATS
 		if (texture->type == TEXTURETYPE_FLAT)
+		{
+			if (W_NeedPaletteRemap())
+				R_DoPaletteRemapFlat(pdata, lumplength);
 			realpatch = (softwarepatch_t *)Picture_Convert(PICFMT_FLAT, pdata, PICFMT_DOOMPATCH, 0, NULL, texture->width, texture->height, 0, 0, 0);
+		}
 		else
 #endif
 		{
 			(void)lumplength;
 			dealloc = false;
+			if (W_NeedPaletteRemap())
+				R_DoPaletteRemapPatch(realpatch, lumplength);
 		}
 
 		HWR_DrawTexturePatchInCache(&grtex->mipmap, blockwidth, blockheight, texture, patch, realpatch);
@@ -878,6 +884,11 @@ static void HWR_CacheFlat(GLMipmap_t *grMipmap, lumpnum_t flatlumpnum)
 	W_ReadLump(flatlumpnum, Z_Malloc(W_LumpLength(flatlumpnum),
 		PU_HWRCACHE, &grMipmap->data));
 
+	// palette remapping
+	// have to do this here since the lump isn't cached
+	if (wadfiles[WADFILENUM(flatlumpnum)]->compatmode)
+		R_DoPaletteRemapFlat(grMipmap->data, size);
+
 	flat = grMipmap->data;
 	for (steppy = 0; steppy < size; steppy++)
 		if (flat[steppy] != HWR_PATCHES_CHROMAKEY_COLORINDEX)
diff --git a/src/r_data.c b/src/r_data.c
index 9ac323814..c2ba15c5f 100644
--- a/src/r_data.c
+++ b/src/r_data.c
@@ -292,6 +292,7 @@ void R_ReInitColormaps(UINT16 num, lumpnum_t newencoremap)
 	char colormap[9] = "COLORMAP";
 	lumpnum_t lump;
 	const lumpnum_t basecolormaplump = W_GetNumForName(colormap);
+	boolean remap = false;
 	if (num > 0 && num <= 10000)
 		snprintf(colormap, 8, "CLM%04u", num-1);
 
@@ -301,6 +302,7 @@ void R_ReInitColormaps(UINT16 num, lumpnum_t newencoremap)
 		lump = basecolormaplump;
 	else
 	{
+		remap = wadfiles[WADFILENUM(lump)]->compatmode;
 		if (W_LumpLength(lump) != W_LumpLength(basecolormaplump))
 		{
 			CONS_Alert(CONS_WARNING, "%s lump size does not match COLORMAP, results may be unexpected.\n", colormap);
@@ -309,6 +311,19 @@ void R_ReInitColormaps(UINT16 num, lumpnum_t newencoremap)
 
 	W_ReadLumpHeader(lump, colormaps, W_LumpLength(basecolormaplump), 0U);
 
+	if (remap)
+	{
+		lighttable_t *copy = malloc(W_LumpLength(basecolormaplump));
+		memcpy(copy, colormaps, W_LumpLength(basecolormaplump));
+		for (int p = 0; p < LIGHTLEVELS; p++)
+		{
+			for (int i = 0; i < 256; i++)
+				// 2.2 colormap index to 2.1, then 2.1 palette color to 2.2
+				colormaps[p*256 + i] = R_GetPaletteRemap(copy[p*256 + R_InvPaletteRemap(i)]);
+		}
+		free(copy);
+	}
+
 	// Encore mode.
 	if (newencoremap != LUMPERROR)
 	{
@@ -320,6 +335,16 @@ void R_ReInitColormaps(UINT16 num, lumpnum_t newencoremap)
 		colormap_p = colormap_p2 = colormaps;
 		colormap_p += COLORMAP_REMAPOFFSET;
 
+		remap = wadfiles[WADFILENUM(newencoremap)]->compatmode;
+		if (remap)
+		{
+			UINT8 *copy = malloc(256);
+			memcpy(copy, encoremap, 256);
+			for (i = 0; i < 256; i++)
+				encoremap[i] = R_GetPaletteRemap(copy[R_InvPaletteRemap(i)]);
+			free(copy);
+		}
+
 		for (p = 0; p < LIGHTLEVELS; p++)
 		{
 			for (i = 0; i < 256; i++)
diff --git a/src/r_draw.cpp b/src/r_draw.cpp
index 588c1cb99..5d6591b02 100644
--- a/src/r_draw.cpp
+++ b/src/r_draw.cpp
@@ -108,6 +108,7 @@ UINT8 r8_flatcolor;
 
 UINT8 *transtables; // translucency tables
 UINT8 *blendtables[NUMBLENDMAPS];
+UINT8 *palremap, *invremap;
 
 /**	\brief R_DrawTransColumn uses this
 */
@@ -348,6 +349,24 @@ UINT8 *R_GetBlendTable(int style, INT32 alphalevel)
 	return transtables + (ClipTransLevel(alphalevel) << FF_TRANSSHIFT);
 }
 
+void R_InitPaletteRemap(void)
+{
+	palremap = static_cast<UINT8 *>(Z_Malloc(256, PU_STATIC, NULL));
+	invremap = static_cast<UINT8 *>(Z_Malloc(256, PU_STATIC, NULL));
+	W_ReadLump(W_GetNumForName("PALREMAP"), palremap);
+	W_ReadLump(W_GetNumForName("INVREMAP"), invremap);
+}
+
+UINT8 R_GetPaletteRemap(UINT8 color)
+{
+	return palremap[color];
+}
+
+UINT8 R_InvPaletteRemap(UINT8 color)
+{
+	return invremap[color];
+}
+
 /**	\brief	Retrieves a translation colormap from the cache.
 
 	\param	skinnum	number of skin, TC_DEFAULT or TC_BOSS
diff --git a/src/r_draw.h b/src/r_draw.h
index a7bc1bec8..0712ab852 100644
--- a/src/r_draw.h
+++ b/src/r_draw.h
@@ -167,11 +167,16 @@ enum
 
 extern UINT8 *blendtables[NUMBLENDMAPS];
 
+extern UINT8 *palremap, *invremap;
+
 void R_InitTranslucencyTables(void);
 void R_GenerateBlendTables(void);
+void R_InitPaletteRemap(void);
 
 UINT8 *R_GetTranslucencyTable(INT32 alphalevel);
 UINT8 *R_GetBlendTable(int style, INT32 alphalevel);
+UINT8 R_GetPaletteRemap(UINT8 color);
+UINT8 R_InvPaletteRemap(UINT8 color);
 
 // Color ramp modification should force a recache
 extern UINT8 skincolor_modified[];
diff --git a/src/r_textures.c b/src/r_textures.c
index 9aa7f68ea..d7e6b77bd 100644
--- a/src/r_textures.c
+++ b/src/r_textures.c
@@ -295,6 +295,28 @@ static boolean R_CheckTextureLumpLength(texture_t *texture, size_t patch)
 	return true;
 }
 
+void R_DoPaletteRemapPatch(softwarepatch_t *patch, size_t size)
+{
+	(void)size;
+	UINT32 *columns = patch->columnofs;
+	for (INT16 i = 0; i < patch->width; i++)
+	{
+		UINT8 *p = (UINT8 *)patch + columns[i];
+		while (*p++ != 0xff) // topdelta
+		{
+			int length = *p++;
+			for (length += 2; length--; p++)
+				*p = R_GetPaletteRemap(*p);
+		}
+	}
+}
+
+void R_DoPaletteRemapFlat(UINT8 *data, size_t size)
+{
+	for (size_t i = 0; i < size; i++, data++)
+		*data = R_GetPaletteRemap(*data);
+}
+
 //
 // R_GenerateTexture
 //
@@ -363,6 +385,9 @@ UINT8 *R_GenerateTexture(size_t texnum)
 			goto multipatch;
 #endif
 
+		if (W_NeedPaletteRemap())
+			R_DoPaletteRemapPatch(realpatch, lumplength);
+
 		// Check the patch for holes.
 		if (texture->width > SHORT(realpatch->width) || texture->height > SHORT(realpatch->height))
 			holey = true;
@@ -458,12 +483,18 @@ UINT8 *R_GenerateTexture(size_t texnum)
 #endif
 #ifdef WALLFLATS
 		if (texture->type == TEXTURETYPE_FLAT)
+		{
+			if (W_NeedPaletteRemap())
+				R_DoPaletteRemapFlat(pdata, lumplength);
 			realpatch = (softwarepatch_t *)Picture_Convert(PICFMT_FLAT, pdata, PICFMT_DOOMPATCH, 0, NULL, texture->width, texture->height, 0, 0, 0);
+		}
 		else
 #endif
 		{
 			(void)lumplength;
 			dealloc = false;
+			if (W_NeedPaletteRemap())
+				R_DoPaletteRemapPatch(realpatch, lumplength);
 		}
 
 		x1 = patch->originx;
@@ -893,7 +924,10 @@ UINT8 *R_GetBrightmapColumn(fixed_t tex, INT32 col)
 
 void *R_GetFlat(lumpnum_t flatlumpnum)
 {
-	return W_CacheLumpNum(flatlumpnum, PU_LEVEL);
+	void *flat = W_CacheLumpNum(flatlumpnum, PU_LEVEL);
+	if (W_NeedPaletteRemap())
+		R_DoPaletteRemapFlat(flat, W_LumpLength(flatlumpnum));
+	return flat;
 }
 
 //
diff --git a/src/r_textures.h b/src/r_textures.h
index 2559766f6..44f1b3d61 100644
--- a/src/r_textures.h
+++ b/src/r_textures.h
@@ -94,6 +94,8 @@ boolean R_TextureHasBrightmap(INT32 texnum);
 boolean R_TextureCanRemap(INT32 texnum);
 void R_CheckTextureCache(INT32 tex);
 void R_ClearTextureNumCache(boolean btell);
+void R_DoPaletteRemapPatch(softwarepatch_t *patch, size_t size);
+void R_DoPaletteRemapFlat(UINT8 *data, size_t size);
 
 // Retrieve texture data.
 void *R_GetLevelFlat(drawspandata_t* ds, levelflat_t *levelflat);
diff --git a/src/v_video.c b/src/v_video.c
index ae032b7bb..9f53cb1de 100644
--- a/src/v_video.c
+++ b/src/v_video.c
@@ -312,6 +312,9 @@ UINT32 V_GammaCorrect(UINT32 input, double power)
 static void LoadPalette(const char *lumpname)
 {
 	lumpnum_t lumpnum = W_GetNumForName(lumpname);
+	boolean remap = wadfiles[WADFILENUM(lumpnum)]->compatmode;
+	if (remap) // if we're remapping, start with PLAYPAL as a base
+		lumpnum = W_GetNumForName("PLAYPAL");
 	size_t i, palsize = W_LumpLength(lumpnum)/3;
 	UINT8 *pal;
 
@@ -329,26 +332,51 @@ static void LoadPalette(const char *lumpname)
 		pLocalPalette = pMasterPalette;
 	pGammaCorrectedPalette = Z_Malloc(sizeof (*pGammaCorrectedPalette)*palsize, PU_STATIC, NULL);
 
+	// load the new palette (or PLAYPAL if remapping)
 	pal = W_CacheLumpNum(lumpnum, PU_CACHE);
 	for (i = 0; i < palsize; i++)
 	{
-		RGBA_t pGCP;
 		pMasterPalette[i].s.red = *pal++;
 		pMasterPalette[i].s.green = *pal++;
 		pMasterPalette[i].s.blue = *pal++;
 		pMasterPalette[i].s.alpha = 0xFF;
+	}
 
+	RGBA_t *target = NULL;
+	if (remap)
+	{
+		// get the target palette we want
+		lumpnum = W_GetNumForName(lumpname);
+		pal = W_CacheLumpNum(lumpnum, PU_CACHE);
+		palsize = W_LumpLength(lumpnum)/3;
+		target = malloc(sizeof (*target)*palsize);
+		for (i = 0; i < palsize; i++)
+		{
+			target[i].s.red = *pal++;
+			target[i].s.green = *pal++;
+			target[i].s.blue = *pal++;
+			target[i].s.alpha = 0xFF;
+		}
+	}
+
+	for (i = 0; i < palsize; i++)
+	{
+		if (remap) // now map the target palette (2.1) to master palette (2.2)
+			pMasterPalette[i].rgba = target[R_InvPaletteRemap(i)].rgba;
 		pGammaCorrectedPalette[i].rgba = V_GammaDecode(pMasterPalette[i].rgba);
 
 		if (!Cubeapply)
 			continue;
 		
 		// Short hand this so its easier to type
-		pGCP = pGammaCorrectedPalette[i];
+		RGBA_t pGCP = pGammaCorrectedPalette[i];
 
 		V_CubeApply(&pGCP.s.red,&pGCP.s.green,&pGCP.s.blue);
 		pLocalPalette[i].rgba = V_GammaEncode(pGammaCorrectedPalette[i].rgba);
 	}
+
+	if (remap)
+		free(target);
 }
 
 void V_CubeApply(UINT8 *red, UINT8 *green, UINT8 *blue)
diff --git a/src/w_wad.c b/src/w_wad.c
index e9f346746..49efb9cf2 100644
--- a/src/w_wad.c
+++ b/src/w_wad.c
@@ -1485,6 +1485,18 @@ void zerr(int ret)
 }
 #endif
 
+// only do palette swapping ONCE, when caching the whole patch
+// dumb hack but it works!
+static boolean dopalswap = false;
+
+// returns true if the most recently loaded lump is freshly cached, AND comes from a compatmode file
+boolean W_NeedPaletteRemap(void)
+{
+	boolean r = dopalswap;
+	dopalswap = false;
+	return r;
+}
+
 /** Reads bytes from the head of a lump.
   * Note: If the lump is compressed, the whole thing has to be read anyway.
   *
@@ -1673,11 +1685,15 @@ void *W_CacheLumpNumPwad(UINT16 wad, UINT16 lump, INT32 tag)
 	lumpcache = wadfiles[wad]->lumpcache;
 	if (!lumpcache[lump])
 	{
+		dopalswap = wadfiles[wad]->compatmode;
 		void *ptr = Z_Malloc(W_LumpLengthPwad(wad, lump), tag, &lumpcache[lump]);
 		W_ReadLumpHeaderPwad(wad, lump, ptr, 0, 0);  // read the lump in full
 	}
 	else
+	{
+		dopalswap = false;
 		Z_ChangeTag(lumpcache[lump], tag);
+	}
 
 	return lumpcache[lump];
 }
@@ -1819,6 +1835,10 @@ void *W_CacheSoftwarePatchNumPwad(UINT16 wad, UINT16 lump, INT32 tag)
 			ptr = Picture_PNGConvert((UINT8 *)lumpdata, PICFMT_DOOMPATCH, NULL, NULL, NULL, NULL, len, &len, 0);
 #endif
 
+		// we already know this is a patch, do palette remapping here
+		if (wadfiles[wad]->compatmode)
+			R_DoPaletteRemapPatch(ptr, wadfiles[wad]->lumpinfo[lump].size);
+
 		dest = Z_Calloc(sizeof(patch_t), tag, &lumpcache[lump]);
 		Patch_Create(ptr, len, dest);
 
diff --git a/src/w_wad.h b/src/w_wad.h
index 27b7174ac..445fe56bc 100644
--- a/src/w_wad.h
+++ b/src/w_wad.h
@@ -196,6 +196,7 @@ void *W_CacheLumpNumForce(lumpnum_t lumpnum, INT32 tag);
 
 boolean W_IsLumpCached(lumpnum_t lump, void *ptr);
 boolean W_IsPatchCached(lumpnum_t lump, void *ptr);
+boolean W_NeedPaletteRemap(void);
 
 void *W_CacheLumpName(const char *name, INT32 tag);
 void *W_CachePatchName(const char *name, INT32 tag);