diff --git a/channels/audin/client/audin_main.c b/channels/audin/client/audin_main.c
index ea552d40c..f9cead2d8 100644
--- a/channels/audin/client/audin_main.c
+++ b/channels/audin/client/audin_main.c
@@ -416,22 +416,24 @@ static BOOL audin_open_device(AUDIN_PLUGIN* audin, AUDIN_CHANNEL_CALLBACK* callb
 	if (!supported)
 	{
 		/* Default sample rates supported by most backends. */
-		const UINT32 samplerates[] = { 96000, 48000, 44100, 22050 };
+		const UINT32 samplerates[] = { format.nSamplesPerSec, 96000, 48000, 44100, 22050 };
 		BOOL test = FALSE;
+		size_t x;
 
 		format.wFormatTag = WAVE_FORMAT_PCM;
 		format.wBitsPerSample = 16;
-		test = IFCALLRESULT(FALSE, audin->device->FormatSupported, audin->device, &format);
-		if (!test)
+		format.cbSize = 0;
+		for (x = 0; x < ARRAYSIZE(samplerates); x++)
 		{
-			size_t x;
-			for (x = 0; x < ARRAYSIZE(samplerates); x++)
-			{
-				format.nSamplesPerSec = samplerates[x];
-				test = IFCALLRESULT(FALSE, audin->device->FormatSupported, audin->device, &format);
-				if (test)
-					break;
-			}
+			format.nSamplesPerSec = samplerates[x];
+			format.nChannels = audin->format->nChannels;
+			test = IFCALLRESULT(FALSE, audin->device->FormatSupported, audin->device, &format);
+			if (test)
+				break;
+			format.nChannels = 3 - format.nChannels;
+			test = IFCALLRESULT(FALSE, audin->device->FormatSupported, audin->device, &format);
+			if (test)
+				break;
 		}
 		if (!test)
 			return FALSE;
@@ -447,7 +449,7 @@ static BOOL audin_open_device(AUDIN_PLUGIN* audin, AUDIN_CHANNEL_CALLBACK* callb
 
 	if (!supported)
 	{
-		if (!freerdp_dsp_context_reset(audin->dsp_context, audin->format))
+		if (!freerdp_dsp_context_reset(audin->dsp_context, audin->format, audin->FramesPerPacket))
 			return FALSE;
 	}
 
@@ -943,7 +945,7 @@ BOOL audin_process_addin_args(AUDIN_PLUGIN* audin, const ADDIN_ARGV* args)
 		{
 			unsigned long val = strtoul(arg->Value, NULL, 0);
 
-			if ((errno != 0) || (val > UINT16_MAX))
+			if ((errno != 0) || (val < UINT16_MAX))
 				audin->fixed_format->nChannels = val;
 		}
 		CommandLineSwitchDefault(arg)
diff --git a/channels/audin/client/mac/audin_mac.m b/channels/audin/client/mac/audin_mac.m
index 48dce2695..a974837f1 100644
--- a/channels/audin/client/mac/audin_mac.m
+++ b/channels/audin/client/mac/audin_mac.m
@@ -120,6 +120,9 @@ static BOOL audin_mac_format_supported(IAudinDevice *device, const AUDIO_FORMAT
 	if (device == NULL || format == NULL)
 		return FALSE;
 
+	if (format->nChannels != 2)
+		return FALSE;
+
 	req_fmt = audin_mac_get_format(format);
 
 	if (req_fmt == 0)
diff --git a/channels/audin/server/audin.c b/channels/audin/server/audin.c
index 13e4cf754..10d61ed26 100644
--- a/channels/audin/server/audin.c
+++ b/channels/audin/server/audin.c
@@ -86,7 +86,7 @@ static UINT audin_server_select_format(audin_server_context* context, size_t cli
 	context->selected_client_format = (SSIZE_T)client_format_index;
 
 	if (!freerdp_dsp_context_reset(audin->dsp_context,
-	                               &audin->context.client_formats[client_format_index]))
+	                               &audin->context.client_formats[client_format_index], 0u))
 	{
 		WLog_ERR(TAG, "Failed to reset dsp context format!");
 		return ERROR_INTERNAL_ERROR;
diff --git a/channels/rdpsnd/client/mac/rdpsnd_mac.m b/channels/rdpsnd/client/mac/rdpsnd_mac.m
index 8a01d9b73..d8c4da8d6 100644
--- a/channels/rdpsnd/client/mac/rdpsnd_mac.m
+++ b/channels/rdpsnd/client/mac/rdpsnd_mac.m
@@ -237,6 +237,9 @@ static BOOL rdpsnd_mac_format_supported(rdpsndDevicePlugin *device, const AUDIO_
 		case WAVE_FORMAT_PCM:
 			if (format->wBitsPerSample != 16)
 				return FALSE;
+
+			if (format->nChannels != 2)
+				return FALSE;
 			return TRUE;
 
 		default:
diff --git a/channels/rdpsnd/client/rdpsnd_main.c b/channels/rdpsnd/client/rdpsnd_main.c
index be48c420c..95fc2cd6e 100644
--- a/channels/rdpsnd/client/rdpsnd_main.c
+++ b/channels/rdpsnd/client/rdpsnd_main.c
@@ -407,7 +407,7 @@ static BOOL rdpsnd_ensure_device_is_open(rdpsndPlugin* rdpsnd, UINT32 wFormatNo,
 
 		if (!supported)
 		{
-			if (!freerdp_dsp_context_reset(rdpsnd->dsp_context, format))
+			if (!freerdp_dsp_context_reset(rdpsnd->dsp_context, format, 0u))
 				return FALSE;
 		}
 
diff --git a/channels/rdpsnd/server/rdpsnd_main.c b/channels/rdpsnd/server/rdpsnd_main.c
index 8e72a76cf..a52f5d72d 100644
--- a/channels/rdpsnd/server/rdpsnd_main.c
+++ b/channels/rdpsnd/server/rdpsnd_main.c
@@ -365,7 +365,7 @@ static UINT rdpsnd_server_select_format(RdpsndServerContext* context, UINT16 cli
 		context->priv->out_buffer_size = out_buffer_size;
 	}
 
-	freerdp_dsp_context_reset(context->priv->dsp_context, format);
+	freerdp_dsp_context_reset(context->priv->dsp_context, format, 0u);
 out:
 	LeaveCriticalSection(&context->priv->lock);
 	return error;
diff --git a/include/freerdp/codec/dsp.h b/include/freerdp/codec/dsp.h
index 7776885ff..779966169 100644
--- a/include/freerdp/codec/dsp.h
+++ b/include/freerdp/codec/dsp.h
@@ -40,7 +40,8 @@ extern "C"
 	                                    const BYTE* data, size_t length, wStream* out);
 	FREERDP_API void freerdp_dsp_context_free(FREERDP_DSP_CONTEXT* context);
 	FREERDP_API BOOL freerdp_dsp_context_reset(FREERDP_DSP_CONTEXT* context,
-	                                           const AUDIO_FORMAT* targetFormat);
+	                                           const AUDIO_FORMAT* targetFormat,
+	                                           UINT32 FramesPerPacket);
 
 #ifdef __cplusplus
 }
diff --git a/libfreerdp/codec/dsp.c b/libfreerdp/codec/dsp.c
index 9c5d8f7ec..29f8ace79 100644
--- a/libfreerdp/codec/dsp.c
+++ b/libfreerdp/codec/dsp.c
@@ -64,6 +64,7 @@
 union _ADPCM {
 	struct
 	{
+		size_t packet_size;
 		INT16 last_sample[2];
 		INT16 last_step[2];
 	} ima;
@@ -84,8 +85,9 @@ struct _FREERDP_DSP_CONTEXT
 	ADPCM adpcm;
 	AUDIO_FORMAT format;
 
-	wStream* buffer;
+	wStream* channelmix;
 	wStream* resample;
+	wStream* buffer;
 
 #if defined(WITH_GSM)
 	gsm gsm;
@@ -136,7 +138,7 @@ static BOOL freerdp_dsp_channel_mix(FREERDP_DSP_CONTEXT* context, const BYTE* sr
 		return FALSE;
 
 	bpp = srcFormat->wBitsPerSample > 8 ? 2 : 1;
-	samples = size / bpp;
+	samples = size / bpp / srcFormat->nChannels;
 
 	if (context->format.nChannels == srcFormat->nChannels)
 	{
@@ -145,7 +147,7 @@ static BOOL freerdp_dsp_channel_mix(FREERDP_DSP_CONTEXT* context, const BYTE* sr
 		return TRUE;
 	}
 
-	Stream_SetPosition(context->buffer, 0);
+	Stream_SetPosition(context->channelmix, 0);
 
 	/* Destination has more channels than source */
 	if (context->format.nChannels > srcFormat->nChannels)
@@ -153,21 +155,21 @@ static BOOL freerdp_dsp_channel_mix(FREERDP_DSP_CONTEXT* context, const BYTE* sr
 		switch (srcFormat->nChannels)
 		{
 			case 1:
-				if (!Stream_EnsureCapacity(context->buffer, size * 2))
+				if (!Stream_EnsureCapacity(context->channelmix, size * 2))
 					return FALSE;
 
 				for (x = 0; x < samples; x++)
 				{
 					for (y = 0; y < bpp; y++)
-						Stream_Write_UINT8(context->buffer, src[x * bpp + y]);
+						Stream_Write_UINT8(context->channelmix, src[x * bpp + y]);
 
 					for (y = 0; y < bpp; y++)
-						Stream_Write_UINT8(context->buffer, src[x * bpp + y]);
+						Stream_Write_UINT8(context->channelmix, src[x * bpp + y]);
 				}
 
-				Stream_SealLength(context->buffer);
-				*data = Stream_Buffer(context->buffer);
-				*length = Stream_Length(context->buffer);
+				Stream_SealLength(context->channelmix);
+				*data = Stream_Buffer(context->channelmix);
+				*length = Stream_Length(context->channelmix);
 				return TRUE;
 
 			case 2:  /* We only support stereo, so we can not handle this case. */
@@ -180,7 +182,7 @@ static BOOL freerdp_dsp_channel_mix(FREERDP_DSP_CONTEXT* context, const BYTE* sr
 	switch (srcFormat->nChannels)
 	{
 		case 2:
-			if (!Stream_EnsureCapacity(context->buffer, size / 2))
+			if (!Stream_EnsureCapacity(context->channelmix, size / 2))
 				return FALSE;
 
 			/* Simply drop second channel.
@@ -188,12 +190,12 @@ static BOOL freerdp_dsp_channel_mix(FREERDP_DSP_CONTEXT* context, const BYTE* sr
 			for (x = 0; x < samples; x++)
 			{
 				for (y = 0; y < bpp; y++)
-					Stream_Write_UINT8(context->buffer, src[2 * x * bpp + y]);
+					Stream_Write_UINT8(context->channelmix, src[2 * x * bpp + y]);
 			}
 
-			Stream_SealLength(context->buffer);
-			*data = Stream_Buffer(context->buffer);
-			*length = Stream_Length(context->buffer);
+			Stream_SealLength(context->channelmix);
+			*data = Stream_Buffer(context->channelmix);
+			*length = Stream_Length(context->channelmix);
 			return TRUE;
 
 		case 1:  /* Invalid, do we want to use a 0 channel sound? */
@@ -238,9 +240,14 @@ static BOOL freerdp_dsp_resample(FREERDP_DSP_CONTEXT* context, const BYTE* src,
 	/* We want to ignore differences of source and destination format. */
 	format = *srcFormat;
 	format.wFormatTag = WAVE_FORMAT_UNKNOWN;
+	format.wBitsPerSample = 0;
 
 	if (audio_format_compatible(&format, &context->format))
+	{
+		*data = src;
+		*length = size;
 		return TRUE;
+	}
 
 #if defined(WITH_SOXR)
 	sbytes = srcChannels * srcBytesPerFrame;
@@ -261,10 +268,6 @@ static BOOL freerdp_dsp_resample(FREERDP_DSP_CONTEXT* context, const BYTE* src,
 	*length = Stream_Length(context->resample);
 	return (error == 0) ? TRUE : FALSE;
 #else
-	WINPR_UNUSED(src);
-	WINPR_UNUSED(size);
-	WINPR_UNUSED(data);
-	WINPR_UNUSED(length);
 	WLog_ERR(TAG, "Missing resample support, recompile -DWITH_SOXR=ON or -DWITH_DSP_FFMPEG=ON");
 	return FALSE;
 #endif
@@ -720,10 +723,11 @@ static BOOL freerdp_dsp_encode_ima_adpcm(FREERDP_DSP_CONTEXT* context, const BYT
 	if (!Stream_EnsureRemainingCapacity(out, size))
 		return FALSE;
 
-	start = dst = Stream_Pointer(out);
+	start = Stream_Buffer(context->buffer);
+	dst = Stream_Pointer(context->buffer);
 	align = (context->format.nChannels > 1) ? 32 : 4;
 
-	while (size > align)
+	while (size >= align)
 	{
 		if ((dst - start) % context->format.nBlockAlign == 0)
 		{
@@ -768,9 +772,15 @@ static BOOL freerdp_dsp_encode_ima_adpcm(FREERDP_DSP_CONTEXT* context, const BYT
 			*dst++ = encoded;
 			size -= 4;
 		}
+
+		if (dst - start == context->adpcm.ima.packet_size)
+		{
+			Stream_Write(out, start, context->adpcm.ima.packet_size);
+			dst = Stream_Buffer(context->buffer);
+		}
 	}
 
-	Stream_SetPointer(out, dst);
+	Stream_SetPointer(context->buffer, dst);
 	return TRUE;
 }
 
@@ -1025,9 +1035,9 @@ FREERDP_DSP_CONTEXT* freerdp_dsp_context_new(BOOL encoder)
 	if (!context)
 		return NULL;
 
-	context->buffer = Stream_New(NULL, 4096);
+	context->channelmix = Stream_New(NULL, 4096);
 
-	if (!context->buffer)
+	if (!context->channelmix)
 		goto fail;
 
 	context->resample = Stream_New(NULL, 4096);
@@ -1035,6 +1045,11 @@ FREERDP_DSP_CONTEXT* freerdp_dsp_context_new(BOOL encoder)
 	if (!context->resample)
 		goto fail;
 
+	context->buffer = Stream_New(NULL, 4096);
+
+	if (!context->buffer)
+		goto fail;
+
 	context->encoder = encoder;
 #if defined(WITH_GSM)
 	context->gsm = gsm_create();
@@ -1095,8 +1110,9 @@ void freerdp_dsp_context_free(FREERDP_DSP_CONTEXT* context)
 
 	if (context)
 	{
-		Stream_Free(context->buffer, TRUE);
+		Stream_Free(context->channelmix, TRUE);
 		Stream_Free(context->resample, TRUE);
+		Stream_Free(context->buffer, TRUE);
 #if defined(WITH_GSM)
 		gsm_destroy(context->gsm);
 #endif
@@ -1254,6 +1270,7 @@ BOOL freerdp_dsp_supports_format(const AUDIO_FORMAT* format, BOOL encode)
 #if defined(WITH_DSP_EXPERIMENTAL)
 
 		case WAVE_FORMAT_ADPCM:
+			return FALSE;
 		case WAVE_FORMAT_DVI_ADPCM:
 			return TRUE;
 #endif
@@ -1297,7 +1314,8 @@ BOOL freerdp_dsp_supports_format(const AUDIO_FORMAT* format, BOOL encode)
 #endif
 }
 
-BOOL freerdp_dsp_context_reset(FREERDP_DSP_CONTEXT* context, const AUDIO_FORMAT* targetFormat)
+BOOL freerdp_dsp_context_reset(FREERDP_DSP_CONTEXT* context, const AUDIO_FORMAT* targetFormat,
+                               UINT32 FramesPerPacket)
 {
 #if defined(WITH_DSP_FFMPEG)
 	return freerdp_dsp_ffmpeg_context_reset(context, targetFormat);
@@ -1307,6 +1325,22 @@ BOOL freerdp_dsp_context_reset(FREERDP_DSP_CONTEXT* context, const AUDIO_FORMAT*
 		return FALSE;
 
 	context->format = *targetFormat;
+
+	if (context->format.wFormatTag == WAVE_FORMAT_DVI_ADPCM)
+	{
+		size_t min_frame_data =
+		    context->format.wBitsPerSample * context->format.nChannels * FramesPerPacket * 1ULL;
+		size_t data_per_block = (context->format.nBlockAlign - 4 * context->format.nChannels) * 8;
+		size_t nb_block_per_packet = min_frame_data / data_per_block;
+
+		if (min_frame_data % data_per_block)
+			nb_block_per_packet++;
+
+		context->adpcm.ima.packet_size = nb_block_per_packet * context->format.nBlockAlign;
+		Stream_EnsureCapacity(context->buffer, context->adpcm.ima.packet_size);
+		Stream_SetPosition(context->buffer, 0);
+	}
+
 #if defined(WITH_FAAD2)
 	context->faadSetup = FALSE;
 #endif