diff --git a/spa/plugins/audiomixer/meson.build b/spa/plugins/audiomixer/meson.build
index c6aa5f2b3..4312566f8 100644
--- a/spa/plugins/audiomixer/meson.build
+++ b/spa/plugins/audiomixer/meson.build
@@ -35,6 +35,16 @@ if have_sse2
 	simd_cargs += ['-DHAVE_SSE2']
 	simd_dependencies += audiomixer_sse2
 endif
+if have_avx and have_fma
+	audiomixer_avx = static_library('audiomixer_avx',
+		['mix-ops-avx.c'],
+		c_args : [avx_args, fma_args, '-O3', '-DHAVE_AVX', '-DHAVE_FMA'],
+		include_directories : [spa_inc],
+		install : false
+	)
+	simd_cargs += ['-DHAVE_AVX', '-DHAVE_FMA']
+	simd_dependencies += audiomixer_avx
+endif
 
 audiomixerlib = shared_library('spa-audiomixer',
                           audiomixer_sources,
diff --git a/spa/plugins/audiomixer/mix-ops-avx.c b/spa/plugins/audiomixer/mix-ops-avx.c
new file mode 100644
index 000000000..7b995d67d
--- /dev/null
+++ b/spa/plugins/audiomixer/mix-ops-avx.c
@@ -0,0 +1,135 @@
+/* Spa
+ *
+ * Copyright © 2019 Wim Taymans
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include <string.h>
+#include <stdio.h>
+#include <math.h>
+
+#include <spa/utils/defs.h>
+
+#include "mix-ops.h"
+
+#include <immintrin.h>
+
+static inline void mix_4(float * dst,
+		const float * SPA_RESTRICT src0,
+		const float * SPA_RESTRICT src1,
+		const float * SPA_RESTRICT src2,
+		uint32_t n_samples)
+{
+	uint32_t n, unrolled;
+
+	if (SPA_IS_ALIGNED(src0, 32) &&
+	    SPA_IS_ALIGNED(src1, 32) &&
+	    SPA_IS_ALIGNED(src2, 32) &&
+	    SPA_IS_ALIGNED(dst, 32))
+		unrolled = n_samples & ~15;
+	else
+		unrolled = 0;
+
+	for (n = 0; n < unrolled; n += 16) {
+		__m256 in1[4], in2[4];
+
+		in1[0] = _mm256_load_ps(&dst[n + 0]);
+		in2[0] = _mm256_load_ps(&dst[n + 8]);
+		in1[1] = _mm256_load_ps(&src0[n + 0]);
+		in2[1] = _mm256_load_ps(&src0[n + 8]);
+		in1[2] = _mm256_load_ps(&src1[n + 0]);
+		in2[2] = _mm256_load_ps(&src1[n + 8]);
+		in1[3] = _mm256_load_ps(&src2[n + 0]);
+		in2[3] = _mm256_load_ps(&src2[n + 8]);
+
+		in1[0] = _mm256_add_ps(in1[0], in1[1]);
+		in2[0] = _mm256_add_ps(in2[0], in2[1]);
+		in1[2] = _mm256_add_ps(in1[2], in1[3]);
+		in2[2] = _mm256_add_ps(in2[2], in2[3]);
+		in1[0] = _mm256_add_ps(in1[0], in1[2]);
+		in2[0] = _mm256_add_ps(in2[0], in2[2]);
+
+		_mm256_store_ps(&dst[n + 0], in1[0]);
+		_mm256_store_ps(&dst[n + 8], in2[0]);
+	}
+	for (; n < n_samples; n++) {
+		__m128 in[4];
+		in[0] = _mm_load_ss(&dst[n]),
+		in[1] = _mm_load_ss(&src0[n]),
+		in[2] = _mm_load_ss(&src1[n]),
+		in[3] = _mm_load_ss(&src2[n]),
+		in[0] = _mm_add_ss(in[0], in[1]);
+		in[2] = _mm_add_ss(in[2], in[3]);
+		in[0] = _mm_add_ss(in[0], in[2]);
+		_mm_store_ss(&dst[n], in[0]);
+	}
+}
+
+
+static inline void mix_2(float * dst, const float * SPA_RESTRICT src, uint32_t n_samples)
+{
+	uint32_t n, unrolled;
+
+	if (SPA_IS_ALIGNED(src, 32) &&
+	    SPA_IS_ALIGNED(dst, 32))
+		unrolled = n_samples & ~15;
+	else
+		unrolled = 0;
+
+	for (n = 0; n < unrolled; n += 16) {
+		__m256 in1[2], in2[2];
+
+		in1[0] = _mm256_load_ps(&dst[n + 0]);
+		in1[1] = _mm256_load_ps(&dst[n + 8]);
+		in2[0] = _mm256_load_ps(&src[n + 0]);
+		in2[1] = _mm256_load_ps(&src[n + 8]);
+
+		in1[0] = _mm256_add_ps(in1[0], in2[0]);
+		in1[1] = _mm256_add_ps(in1[1], in2[1]);
+
+		_mm256_store_ps(&dst[n + 0], in1[0]);
+		_mm256_store_ps(&dst[n + 8], in1[1]);
+	}
+	for (; n < n_samples; n++) {
+		__m128 in1[0], in2[0];
+		in1[0] = _mm_load_ss(&dst[n]),
+		in2[0] = _mm_load_ss(&src[n]),
+		in1[0] = _mm_add_ss(in1[0], in2[0]);
+		_mm_store_ss(&dst[n], in1[0]);
+	}
+}
+
+void
+mix_f32_avx(struct mix_ops *ops, void * SPA_RESTRICT dst, const void * SPA_RESTRICT src[],
+		uint32_t n_src, uint32_t n_samples)
+{
+	uint32_t i;
+
+	if (n_src == 0)
+		memset(dst, 0, n_samples * sizeof(float));
+	else if (dst != src[0])
+		memcpy(dst, src[0], n_samples * sizeof(float));
+
+	for (i = 1; i + 2 < n_src; i += 3)
+		mix_4(dst, src[i], src[i + 1], src[i + 2], n_samples);
+	for (; i < n_src; i++)
+		mix_2(dst, src[i], n_samples);
+}
diff --git a/spa/plugins/audiomixer/mix-ops.c b/spa/plugins/audiomixer/mix-ops.c
index 6f6acbafe..3d1c0f8e5 100644
--- a/spa/plugins/audiomixer/mix-ops.c
+++ b/spa/plugins/audiomixer/mix-ops.c
@@ -46,6 +46,10 @@ struct mix_info {
 static struct mix_info mix_table[] =
 {
 	/* f32 */
+#if defined(HAVE_AVX)
+	{ SPA_AUDIO_FORMAT_F32, 1, SPA_CPU_FLAG_AVX, 4, mix_f32_avx },
+	{ SPA_AUDIO_FORMAT_F32P, 1, SPA_CPU_FLAG_AVX, 4, mix_f32_avx },
+#endif
 #if defined (HAVE_SSE)
 	{ SPA_AUDIO_FORMAT_F32, 1, SPA_CPU_FLAG_SSE, 4, mix_f32_sse },
 	{ SPA_AUDIO_FORMAT_F32P, 1, SPA_CPU_FLAG_SSE, 4, mix_f32_sse },
diff --git a/spa/plugins/audiomixer/mix-ops.h b/spa/plugins/audiomixer/mix-ops.h
index 692fae3bd..0c878eeca 100644
--- a/spa/plugins/audiomixer/mix-ops.h
+++ b/spa/plugins/audiomixer/mix-ops.h
@@ -59,3 +59,6 @@ DEFINE_FUNCTION(f32, sse);
 #if defined(HAVE_SSE2)
 DEFINE_FUNCTION(f64, sse2);
 #endif
+#if defined(HAVE_AVX)
+DEFINE_FUNCTION(f32, avx);
+#endif