Merge pull request #466 from SuslikV/float-to-int16

Fix distortion during mix of audio

Author: ferdnyc

src/FFmpegReader.cpp

@@ -1627,9 +1627,8 @@ void FFmpegReader::ProcessAudioPacket(int64_t requested_frame, int64_t target_fr
 			else
 				partial_frame = true;
 
-			// Add samples for current channel to the frame. Reduce the volume to 98%, to prevent
-			// some louder samples from maxing out at 1.0 (not sure why this happens)
-			f->AddAudio(true, channel_filter, start, iterate_channel_buffer, samples, 0.98f);
+			// Add samples for current channel to the frame.
+			f->AddAudio(true, channel_filter, start, iterate_channel_buffer, samples, 1.0f);
 
 			// Debug output
 			ZmqLogger::Instance()->AppendDebugMethod("FFmpegReader::ProcessAudioPacket (f->AddAudio)", "frame", starting_frame_number, "start", start, "samples", samples, "channel", channel_filter, "partial_frame", partial_frame, "samples_per_frame", samples_per_frame);

src/FFmpegWriter.cpp

@@ -1525,11 +1525,23 @@ void FFmpegWriter::write_audio_packets(bool is_final) {
 			// Calculate total samples
 			total_frame_samples = samples_in_frame * channels_in_frame;
 
-			// Translate audio sample values back to 16 bit integers
-			for (int s = 0; s < total_frame_samples; s++, frame_position++)
-				// Translate sample value and copy into buffer
-				all_queued_samples[frame_position] = int(frame_samples_float[s] * (1 << 15));
-
+			// Translate audio sample values back to 16 bit integers with saturation
+			float valF;
+			int16_t conv;
+			const int16_t max16 = 32767;
+			const int16_t min16 = -32768;
+			for (int s = 0; s < total_frame_samples; s++, frame_position++) {
+				valF = frame_samples_float[s] * (1 << 15);
+				if (valF > max16)
+					conv = max16;
+				else if (valF < min16)
+					conv = min16;
+				else
+					conv = int(valF + 32768.5) - 32768; // +0.5 is for rounding
+
+				// Copy into buffer
+				all_queued_samples[frame_position] = conv;
+			}
 
 			// Deallocate float array
 			delete[] frame_samples_float;

src/FrameMapper.cpp

@@ -785,10 +785,23 @@ void FrameMapper::ResampleMappedAudio(std::shared_ptr<Frame> frame, int64_t orig
 	// Create a new array (to hold all S16 audio samples for the current queued frames)
  	int16_t* frame_samples = (int16_t*) av_malloc(sizeof(int16_t)*total_frame_samples);
 
-	// Translate audio sample values back to 16 bit integers
-	for (int s = 0; s < total_frame_samples; s++)
-		// Translate sample value and copy into buffer
-		frame_samples[s] = int(frame_samples_float[s] * (1 << 15));
+	// Translate audio sample values back to 16 bit integers with saturation
+	float valF;
+	int16_t conv;
+	const int16_t max16 = 32767;
+	const int16_t min16 = -32768;
+	for (int s = 0; s < total_frame_samples; s++) {
+		valF = frame_samples_float[s] * (1 << 15);
+		if (valF > max16)
+			conv = max16;
+		else if (valF < min16)
+			conv = min16;
+		else
+			conv = int(valF + 32768.5) - 32768; // +0.5 is for rounding
+
+		// Copy into buffer
+		frame_samples[s] = conv;
+	}
 
 
 	// Deallocate float array

tests/FFmpegReader_Tests.cpp

@@ -76,8 +76,8 @@ TEST(FFmpegReader_Check_Audio_File)
 	CHECK_CLOSE(0.0f, samples[50], 0.00001);
 	CHECK_CLOSE(0.0f, samples[100], 0.00001);
 	CHECK_CLOSE(0.0f, samples[200], 0.00001);
-	CHECK_CLOSE(0.160781f, samples[230], 0.00001);
-	CHECK_CLOSE(-0.06125f, samples[300], 0.00001);
+	CHECK_CLOSE(0.16406f, samples[230], 0.00001);
+	CHECK_CLOSE(-0.06250f, samples[300], 0.00001);
 
 	// Close reader
 	r.Close();