diff --git a/GPU/Common/FragmentShaderGenerator.cpp b/GPU/Common/FragmentShaderGenerator.cpp
index fda8d53639..4d74670013 100644
--- a/GPU/Common/FragmentShaderGenerator.cpp
+++ b/GPU/Common/FragmentShaderGenerator.cpp
@@ -685,7 +685,7 @@ bool GenerateFragmentShader(const FShaderID &id, char *buffer, const ShaderLangu
 				}
 				WRITE(p, "  vec2 tsize = vec2(textureSize(tex, 0).xy);\n");
 				WRITE(p, "  vec2 fraction;\n");
-				WRITE(p, "  bool bilinear = (u_depal_mask_shift_off_fmt >> 0x2Fu) != 0x0u;\n");
+				WRITE(p, "  bool bilinear = (u_depal_mask_shift_off_fmt >> 0x1Fu) != 0x0u;\n");
 				WRITE(p, "  if (bilinear) {\n");
 				WRITE(p, "    uv_round = uv * tsize - vec2(0.5, 0.5);\n");
 				WRITE(p, "    fraction = fract(uv_round);\n");
@@ -694,6 +694,7 @@ bool GenerateFragmentShader(const FShaderID &id, char *buffer, const ShaderLangu
 				WRITE(p, "    uv_round = uv;\n");
 				WRITE(p, "  }\n");
 
+				// TODO: Use fetch here when possible.
 				p.C("  highp vec4 t = ").SampleTexture2D("tex", "uv_round").C(";\n");
 				p.C("  highp vec4 t1 = ").SampleTexture2DOffset("tex", "uv_round", 1, 0).C(";\n");
 				p.C("  highp vec4 t2 = ").SampleTexture2DOffset("tex", "uv_round", 0, 1).C(";\n");
diff --git a/GPU/Common/FramebufferManagerCommon.cpp b/GPU/Common/FramebufferManagerCommon.cpp
index 211b6843a0..e56122ad26 100644
--- a/GPU/Common/FramebufferManagerCommon.cpp
+++ b/GPU/Common/FramebufferManagerCommon.cpp
@@ -944,7 +944,7 @@ Draw2DPipeline *FramebufferManagerCommon::GetReinterpretPipeline(GEBufferFormat
 
 	Draw2DPipeline *pipeline = reinterpretFromTo_[(int)from][(int)to];
 	if (!pipeline) {
-		pipeline = draw2D_.Create2DPipeline([=](ShaderWriter &shaderWriter) -> Draw2DPipelineInfo {
+		pipeline = draw2D_.Create2DPipeline([from, to](ShaderWriter &shaderWriter) -> Draw2DPipelineInfo {
 			return GenerateReinterpretFragmentShader(shaderWriter, from, to);
 		});
 		reinterpretFromTo_[(int)from][(int)to] = pipeline;
diff --git a/GPU/Common/ShaderUniforms.cpp b/GPU/Common/ShaderUniforms.cpp
index 5c75c79b9a..a382b59f0b 100644
--- a/GPU/Common/ShaderUniforms.cpp
+++ b/GPU/Common/ShaderUniforms.cpp
@@ -193,9 +193,24 @@ void BaseUpdateUniforms(UB_VS_FS_Base *ub, uint64_t dirtyUniforms, bool useBuffe
 		int indexOffset = gstate.getClutIndexStartPos() >> 4;
 		int format = gstate_c.depalTextureFormat;
 		uint32_t val = BytesToUint32(indexMask, indexShift, indexOffset, format);
-		// Poke in a bilinear filter flag in the top bit.
-		if (gstate.isMagnifyFilteringEnabled())
+		// NOTE: This must follow similar logic to TextureCacheCommon::GetSamplingParams -
+		// maybe we can share it somehow.
+		// TOOD: Handle replaced textures.
+		bool bilinear = gstate.isMagnifyFilteringEnabled() && !gstate_c.pixelMapped;
+		switch (g_Config.iTexFiltering) {
+		case TEX_FILTER_FORCE_NEAREST:
+			bilinear = false;
+			break;
+		case TEX_FILTER_FORCE_LINEAR:
+			bilinear = true;
+			break;
+		default:
+			break;
+		}
+		if (bilinear) {
+			// Poke in a bilinear filter flag in the top bit.
 			val |= 0x80000000;
+		}
 		ub->depal_mask_shift_off_fmt = val;
 	}
 }
diff --git a/GPU/Common/TextureCacheCommon.cpp b/GPU/Common/TextureCacheCommon.cpp
index 0bd6347ae4..fa069d030b 100644
--- a/GPU/Common/TextureCacheCommon.cpp
+++ b/GPU/Common/TextureCacheCommon.cpp
@@ -669,7 +669,7 @@ TexCacheEntry *TextureCacheCommon::SetTexture() {
 
 		if (hasClutGPU) {
 			WARN_LOG_N_TIMES(clutUseRender, 5, Log::G3D, "Using texture with dynamic CLUT: texfmt=%d, clutfmt=%d", gstate.getTextureFormat(), gstate.getClutPaletteFormat());
-			entry->status |= TexStatus::CLUT_GPU;
+			entry->status |= TexStatus::CLUT_GPU | TexStatus::CLUT8_INDEXED;
 		}
 
 		if (hasClut && clutRenderAddress_ == 0xFFFFFFFF) {
@@ -691,6 +691,8 @@ TexCacheEntry *TextureCacheCommon::SetTexture() {
 		}
 
 		nextNeedsChange_ = false;
+	} else {
+		// We had an entry already, modify it.
 	}
 
 	// We have to decode it, let's setup the cache entry first.
@@ -704,6 +706,14 @@ TexCacheEntry *TextureCacheCommon::SetTexture() {
 
 	entry->cluthash = cluthash;
 
+	/*
+	if (entry->maxLevel == 0 && (entry->format == GE_TFMT_CLUT8 || entry->format == GE_TFMT_CLUT4) &&
+		!(entry->status & TexStatus::TO_REPLACE) && !(entry->status & TexStatus::TO_SCALE) && !(entry->status & TexStatus::RELIABLE)) {
+		// Experiment with CLUT8 decoding for regular textures.
+		entry->status |= TexStatus::CLUT8_INDEXED;
+	}
+	*/
+
 	gstate_c.curTextureWidth = w;
 	gstate_c.curTextureHeight = h;
 	gstate_c.SetTextureIsVideo((entry->status & TexStatus::VIDEO) != 0);
@@ -720,7 +730,6 @@ TexCacheEntry *TextureCacheCommon::SetTexture() {
 	nextTexture_ = entry;
 	nextFramebufferTexture_ = nullptr;
 	nextNeedsRehash_ = true;
-	// We still need to rebuild, to allocate a texture.  But we'll bail early.
 	nextNeedsRebuild_ = true;
 	return entry;
 }
@@ -2216,16 +2225,16 @@ void TextureCacheCommon::ApplyTexture(bool doBind, bool flatZ) {
 	}
 
 	gstate_c.SetTextureIsVideo((entry->status & TexStatus::VIDEO) != 0);
+	entry->lastFrame = gpuStats.totals.numFlips;
 	if (entry->status & TexStatus::CLUT_GPU) {
+		_dbg_assert_(entry->status & TexStatus::CLUT8_INDEXED);
 		// Special process.
-		ApplyTextureDepal(entry);
-		entry->lastFrame = gpuStats.totals.numFlips;
+		ApplyTextureDepalFramebufferCLUT(entry);
 		gstate_c.SetTextureSolidAlpha(false);
 		gstate_c.SetTextureIs3D(false);
 		gstate_c.SetTextureIsArray(false);
 		gstate_c.SetTextureIsBGRA(false);
 	} else {
-		entry->lastFrame = gpuStats.totals.numFlips;
 		if (doBind) {
 			BindTexture(entry, flatZ);
 		}
@@ -2233,7 +2242,15 @@ void TextureCacheCommon::ApplyTexture(bool doBind, bool flatZ) {
 		gstate_c.SetTextureIs3D((entry->status & TexStatus::IS_3D) != 0);
 		gstate_c.SetTextureIsArray(false);
 		gstate_c.SetTextureIsBGRA((entry->status & TexStatus::BGRA) != 0);
-		gstate_c.SetShaderDepal(ShaderDepalMode::OFF);
+		if (entry->status & TexStatus::CLUT8_INDEXED) {
+			bool smoothedDepal = false;
+			u32 depthUpperBits = 0;
+			ClutTexture clutTexture = clutTextureCache_.GetClutTexture(gstate.getClutPaletteFormat(), clutHash_, clutBuf_);
+			BindAsClutTexture(clutTexture.texture, false);
+			gstate_c.SetShaderDepal(ShaderDepalMode::NORMAL, GE_FORMAT_CLUT8);
+		} else {
+			gstate_c.SetShaderDepal(ShaderDepalMode::OFF);
+		}
 	}
 }
 
@@ -2474,7 +2491,7 @@ void TextureCacheCommon::ApplyTextureFramebuffer(VirtualFramebuffer *framebuffer
 }
 
 // Applies depal to a normal (non-framebuffer) texture, pre-decoded to CLUT8 format.
-void TextureCacheCommon::ApplyTextureDepal(TexCacheEntry *entry) {
+void TextureCacheCommon::ApplyTextureDepalFramebufferCLUT(TexCacheEntry *entry) {
 	uint32_t clutMode = gstate.clutformat & 0xFFFFFF;
 
 	switch (entry->format) {
@@ -2881,7 +2898,7 @@ bool TextureCacheCommon::PrepareBuildTexture(BuildTexturePlan &plan, TexCacheEnt
 	}
 
 	bool canReplace = !isPPGETexture;
-	if (entry->status & TexStatus::CLUT_GPU) {
+	if (entry->status & TexStatus::CLUT8_INDEXED) {
 		_dbg_assert_(entry->format == GE_TFMT_CLUT4 || entry->format == GE_TFMT_CLUT8);
 		plan.decodeToClut8 = true;
 		// We only support 1 mip level when doing CLUT on GPU for now.
@@ -3001,7 +3018,7 @@ void TextureCacheCommon::LoadTextureLevel(TexCacheEntry &entry, uint8_t *data, s
 		if (!gstate_c.Use(GPU_USE_16BIT_FORMATS) || dstFmt == Draw::DataFormat::R8G8B8A8_UNORM) {
 			texDecFlags |= TexDecodeFlags::EXPAND32;
 		}
-		if (entry.status & (TexStatus::CLUT_GPU | TexStatus::CLUT8_INDEXED)) {
+		if (entry.status & TexStatus::CLUT8_INDEXED) {
 			_dbg_assert_(entry.format == GE_TFMT_CLUT4 || entry.format == GE_TFMT_CLUT8);
 			texDecFlags |= TexDecodeFlags::TO_CLUT8;
 		}
@@ -3080,6 +3097,9 @@ std::string TexStatusToString(TexStatus status) {
 	if (status & TexStatus::CLUT_GPU) {
 		result += "CLUT_GPU ";
 	}
+	if (status & TexStatus::CLUT8_INDEXED) {
+		result += "CLUT8_INDEXED ";
+	}
 	if (status & TexStatus::FORCE_REBUILD) {
 		result += "FORCE_REBUILD ";
 	}
diff --git a/GPU/Common/TextureCacheCommon.h b/GPU/Common/TextureCacheCommon.h
index 8804620089..98ccdc724f 100644
--- a/GPU/Common/TextureCacheCommon.h
+++ b/GPU/Common/TextureCacheCommon.h
@@ -404,7 +404,7 @@ protected:
 	void Decimate(TexCacheEntry *exceptThisOne, bool forcePressure);  // forcePressure defaults to false.
 
 	void ApplyTextureFramebuffer(VirtualFramebuffer *framebuffer, GETextureFormat texFormat, RasterChannel channel);
-	void ApplyTextureDepal(TexCacheEntry *entry);
+	void ApplyTextureDepalFramebufferCLUT(TexCacheEntry *entry);
 
 	void HandleTextureChange(TexCacheEntry *const entry, const char *reason, bool initialMatch, bool doDelete);
 	virtual void BuildTexture(TexCacheEntry *const entry) = 0;
diff --git a/GPU/GeConstants.cpp b/GPU/GeConstants.cpp
index 6cb9bfd361..ce0037cd6b 100644
--- a/GPU/GeConstants.cpp
+++ b/GPU/GeConstants.cpp
@@ -287,6 +287,7 @@ const char *GeBufferFormatToString(GEBufferFormat fmt) {
 	case GE_FORMAT_565: return "565";
 	case GE_FORMAT_8888: return "8888";
 	case GE_FORMAT_DEPTH16: return "DEPTH16";
+	case GE_FORMAT_CLUT8: return "CLUT8";
 	default: return "N/A";
 	}
 }
diff --git a/GPU/Vulkan/TextureCacheVulkan.cpp b/GPU/Vulkan/TextureCacheVulkan.cpp
index e0852bbd6c..43f3b82b23 100644
--- a/GPU/Vulkan/TextureCacheVulkan.cpp
+++ b/GPU/Vulkan/TextureCacheVulkan.cpp
@@ -958,7 +958,7 @@ void TextureCacheVulkan::LoadVulkanTextureLevel(TexCacheEntry &entry, uint8_t *w
 	if (!gstate_c.Use(GPU_USE_16BIT_FORMATS) || scaleFactor > 1 || dstFmt == VULKAN_8888_FORMAT) {
 		texDecFlags |= TexDecodeFlags::EXPAND32;
 	}
-	if (entry.status & TexStatus::CLUT_GPU) {
+	if (entry.status & TexStatus::CLUT8_INDEXED) {
 		texDecFlags |= TexDecodeFlags::TO_CLUT8;
 	}