From 3765a9fe35703d97f27898516eec7e51dce799cd Mon Sep 17 00:00:00 2001
From: akallabeth <akallabeth@posteo.net>
Date: Tue, 11 May 2021 11:32:50 +0200
Subject: [PATCH] Fixed AAC encoding for microphone

* Fixed b-frame interval for FFMPEG
* Fixed sample size for FAAC
---
 libfreerdp/CMakeLists.txt     |  4 +++
 libfreerdp/codec/dsp.c        | 46 +++++++++++++++++++----------------
 libfreerdp/codec/dsp_ffmpeg.c |  2 ++
 3 files changed, 31 insertions(+), 21 deletions(-)

diff --git a/libfreerdp/CMakeLists.txt b/libfreerdp/CMakeLists.txt
index e3a410316..ef76d059b 100644
--- a/libfreerdp/CMakeLists.txt
+++ b/libfreerdp/CMakeLists.txt
@@ -115,6 +115,10 @@ foreach(${MODULE_PREFIX}_SUBMODULE ${${MODULE_PREFIX}_SUBMODULES})
     add_subdirectory(${${MODULE_PREFIX}_SUBMODULE})
 endforeach()
 
+if (NOT WITH_DSP_FFMPEG AND NOT WITH_FAAC)
+    message(WARNING "Compiling without WITH_DSP_FFMPEG and WITH_FAAC, AAC encoder support disabled")
+endif ()
+
 ## cmake source properties are only seen by targets in the same CMakeLists.txt
 ## therefore primitives and codecs need to be defined here
 
diff --git a/libfreerdp/codec/dsp.c b/libfreerdp/codec/dsp.c
index a5039cea5..431bfa4e8 100644
--- a/libfreerdp/codec/dsp.c
+++ b/libfreerdp/codec/dsp.c
@@ -531,7 +531,6 @@ static BOOL freerdp_dsp_encode_faac(FREERDP_DSP_CONTEXT* context, const BYTE* sr
                                     wStream* out)
 {
 	const int16_t* inSamples = (const int16_t*)src;
-	int32_t* outSamples;
 	unsigned int bpp;
 	size_t nrSamples, x;
 	int rc;
@@ -539,28 +538,29 @@ static BOOL freerdp_dsp_encode_faac(FREERDP_DSP_CONTEXT* context, const BYTE* sr
 	if (!context || !src || !out)
 		return FALSE;
 
-	bpp = context->format.wBitsPerSample / 8 * context->format.nChannels;
+	bpp = context->format.wBitsPerSample / 8;
 	nrSamples = size / bpp;
 
-	if (!Stream_EnsureCapacity(context->buffer,
-	                           nrSamples * sizeof(int32_t) * context->format.nChannels))
+	if (!Stream_EnsureRemainingCapacity(context->buffer, nrSamples * sizeof(int16_t)))
 		return FALSE;
 
-	if (!Stream_EnsureRemainingCapacity(out, context->faacMaxOutputBytes))
-		return FALSE;
-
-	outSamples = (int32_t*)Stream_Buffer(context->buffer);
-
-	for (x = 0; x < nrSamples * context->format.nChannels; x++)
-		outSamples[x] = inSamples[x];
-
-	rc = faacEncEncode(context->faac, outSamples, nrSamples * context->format.nChannels,
-	                   Stream_Pointer(out), Stream_GetRemainingCapacity(out));
-
-	if (rc < 0)
-		return FALSE;
-	else if (rc > 0)
-		Stream_Seek(out, (size_t)rc);
+	for (x = 0; x < nrSamples; x++)
+	{
+		Stream_Write_INT16(context->buffer, inSamples[x]);
+		if (Stream_GetPosition(context->buffer) / bpp >= context->faacInputSamples)
+		{
+			if (!Stream_EnsureRemainingCapacity(out, context->faacMaxOutputBytes))
+				return FALSE;
+			rc = faacEncEncode(context->faac, (int32_t*)Stream_Buffer(context->buffer),
+			                   context->faacInputSamples, Stream_Pointer(out),
+			                   Stream_GetRemainingCapacity(out));
+			if (rc < 0)
+				return FALSE;
+			if (rc > 0)
+				Stream_Seek(out, (size_t)rc);
+			Stream_SetPosition(context->buffer, 0);
+		}
+	}
 
 	return TRUE;
 }
@@ -1282,7 +1282,7 @@ BOOL freerdp_dsp_supports_format(const AUDIO_FORMAT* format, BOOL encode)
 				return TRUE;
 
 #endif
-#if defined(WITH_FAAC) && defined(WITH_DSP_EXPERIMENTAL)
+#if defined(WITH_FAAC)
 
 			if (encode)
 				return TRUE;
@@ -1326,7 +1326,11 @@ BOOL freerdp_dsp_context_reset(FREERDP_DSP_CONTEXT* context, const AUDIO_FORMAT*
 			return FALSE;
 
 		cfg = faacEncGetCurrentConfiguration(context->faac);
-		cfg->bitRate = 10000;
+		cfg->inputFormat = FAAC_INPUT_16BIT;
+		cfg->outputFormat = 0;
+		cfg->mpegVersion = MPEG4;
+		cfg->useTns = 1;
+		cfg->bandWidth = targetFormat->nAvgBytesPerSec;
 		faacEncSetConfiguration(context->faac, cfg);
 	}
 
diff --git a/libfreerdp/codec/dsp_ffmpeg.c b/libfreerdp/codec/dsp_ffmpeg.c
index 0af3cf4fc..d680bd763 100644
--- a/libfreerdp/codec/dsp_ffmpeg.c
+++ b/libfreerdp/codec/dsp_ffmpeg.c
@@ -270,6 +270,8 @@ static BOOL ffmpeg_open_context(FREERDP_DSP_CONTEXT* context)
 			break;
 	}
 
+	context->context->max_b_frames = 1;
+	context->context->delay = 0;
 	context->context->channels = format->nChannels;
 	context->context->channel_layout = layout;
 	context->context->sample_rate = format->nSamplesPerSec;