diff --git a/GPU/Common/TextureDecoder.cpp b/GPU/Common/TextureDecoder.cpp
index be528411c..8a5f98e30 100644
--- a/GPU/Common/TextureDecoder.cpp
+++ b/GPU/Common/TextureDecoder.cpp
@@ -659,12 +659,16 @@ CheckAlphaResult CheckAlphaABGR1555Basic(const u32 *pixelData, int stride, int w
 }
 
 CheckAlphaResult CheckAlphaRGBA4444Basic(const u32 *pixelData, int stride, int w, int h) {
-#ifdef _M_SSE
 	// Use SSE if aligned to 16 bytes / 8 pixels (usually the case.)
 	if ((w & 7) == 0 && (stride & 7) == 0) {
+#ifdef _M_SSE
 		return CheckAlphaRGBA4444SSE2(pixelData, stride, w, h);
-	}
+#elif PPSSPP_ARCH(ARMV7) || PPSSPP_ARCH(ARM64)
+		if (cpu_info.bNEON) {
+			return CheckAlphaRGBA4444NEON(pixelData, stride, w, h);
+		}
 #endif
+	}
 
 	const u32 *p = pixelData;
 	const int w2 = (w + 1) / 2;
@@ -688,12 +692,16 @@ CheckAlphaResult CheckAlphaRGBA4444Basic(const u32 *pixelData, int stride, int w
 }
 
 CheckAlphaResult CheckAlphaRGBA5551Basic(const u32 *pixelData, int stride, int w, int h) {
-#ifdef _M_SSE
 	// Use SSE if aligned to 16 bytes / 8 pixels (usually the case.)
 	if ((w & 7) == 0 && (stride & 7) == 0) {
+#ifdef _M_SSE
 		return CheckAlphaRGBA5551SSE2(pixelData, stride, w, h);
-	}
+#elif PPSSPP_ARCH(ARMV7) || PPSSPP_ARCH(ARM64)
+		if (cpu_info.bNEON) {
+			return CheckAlphaRGBA5551NEON(pixelData, stride, w, h);
+		}
 #endif
+	}
 
 	const u32 *p = pixelData;
 	const int w2 = (w + 1) / 2;
diff --git a/GPU/Common/TextureDecoderNEON.cpp b/GPU/Common/TextureDecoderNEON.cpp
index 2b98fdfdf..2a4d06572 100644
--- a/GPU/Common/TextureDecoderNEON.cpp
+++ b/GPU/Common/TextureDecoderNEON.cpp
@@ -342,4 +342,48 @@ CheckAlphaResult CheckAlphaABGR1555NEON(const u32 *pixelData, int stride, int w,
 	return CHECKALPHA_FULL;
 }
 
+CheckAlphaResult CheckAlphaRGBA4444NEON(const u32 *pixelData, int stride, int w, int h) {
+	const u16 *p = (const u16 *)pixelData;
+
+	const uint16x8_t mask = vdupq_n_u16((u16)0xF000);
+	uint16x8_t bits = mask;
+	for (int y = 0; y < h; ++y) {
+		for (int i = 0; i < w; i += 8) {
+			const uint16x8_t a = vld1q_u16(&p[i]);
+			bits = vandq_u16(bits, a);
+		}
+
+		uint16x8_t result = veorq_u16(bits, mask);
+		if (VectorIsNonZeroNEON(result)) {
+			return CHECKALPHA_ANY;
+		}
+
+		p += stride;
+	}
+
+	return CHECKALPHA_FULL;
+}
+
+CheckAlphaResult CheckAlphaRGBA5551NEON(const u32 *pixelData, int stride, int w, int h) {
+	const u16 *p = (const u16 *)pixelData;
+
+	const uint16x8_t mask = vdupq_n_u16((u16)0x8000);
+	uint16x8_t bits = mask;
+	for (int y = 0; y < h; ++y) {
+		for (int i = 0; i < w; i += 8) {
+			const uint16x8_t a = vld1q_u16(&p[i]);
+			bits = vandq_u16(bits, a);
+		}
+
+		uint16x8_t result = veorq_u16(bits, mask);
+		if (VectorIsNonZeroNEON(result)) {
+			return CHECKALPHA_ANY;
+		}
+
+		p += stride;
+	}
+
+	return CHECKALPHA_FULL;
+}
+
 #endif
diff --git a/GPU/Common/TextureDecoderNEON.h b/GPU/Common/TextureDecoderNEON.h
index 1471b41d3..02c3a8ec7 100644
--- a/GPU/Common/TextureDecoderNEON.h
+++ b/GPU/Common/TextureDecoderNEON.h
@@ -24,3 +24,5 @@ u32 ReliableHash32NEON(const void *input, size_t len, u32 seed);
 CheckAlphaResult CheckAlphaRGBA8888NEON(const u32 *pixelData, int stride, int w, int h);
 CheckAlphaResult CheckAlphaABGR4444NEON(const u32 *pixelData, int stride, int w, int h);
 CheckAlphaResult CheckAlphaABGR1555NEON(const u32 *pixelData, int stride, int w, int h);
+CheckAlphaResult CheckAlphaRGBA4444NEON(const u32 *pixelData, int stride, int w, int h);
+CheckAlphaResult CheckAlphaRGBA5551NEON(const u32 *pixelData, int stride, int w, int h);