Merge pull request #268 from clone2727/video-rewrite

VideoDecoder upgrade & partial rewrite

8 files changed, 358 insertions, 317 deletions

diff --git a/engines/sci/console.cpp b/engines/sci/console.cpp
index 564bbbbd79..1889d53480 100644
--- a/engines/sci/console.cpp
+++ b/engines/sci/console.cpp
@@ -250,20 +250,18 @@ void Console::postEnter() {
 #endif
 
 		if (_videoFile.hasSuffix(".seq")) {
-			SeqDecoder *seqDecoder = new SeqDecoder();
-			seqDecoder->setFrameDelay(_videoFrameDelay);
-			videoDecoder = seqDecoder;
+			videoDecoder = new SEQDecoder(_videoFrameDelay);
 #ifdef ENABLE_SCI32
 		} else if (_videoFile.hasSuffix(".vmd")) {
-			videoDecoder = new Video::VMDDecoder(g_system->getMixer());
+			videoDecoder = new Video::AdvancedVMDDecoder();
 		} else if (_videoFile.hasSuffix(".rbt")) {
-			videoDecoder = new RobotDecoder(g_system->getMixer(), _engine->getPlatform() == Common::kPlatformMacintosh);
+			videoDecoder = new RobotDecoder(_engine->getPlatform() == Common::kPlatformMacintosh);
 		} else if (_videoFile.hasSuffix(".duk")) {
 			duckMode = true;
-			videoDecoder = new Video::AviDecoder(g_system->getMixer());
+			videoDecoder = new Video::AVIDecoder();
 #endif
 		} else if (_videoFile.hasSuffix(".avi")) {
-			videoDecoder = new Video::AviDecoder(g_system->getMixer());
+			videoDecoder = new Video::AVIDecoder();
 		} else {
 			warning("Unrecognized video type");
 		}
diff --git a/engines/sci/engine/kvideo.cpp b/engines/sci/engine/kvideo.cpp
index cb2a763da9..6bf9aff2fe 100644
--- a/engines/sci/engine/kvideo.cpp
+++ b/engines/sci/engine/kvideo.cpp
@@ -50,6 +50,8 @@ void playVideo(Video::VideoDecoder *videoDecoder, VideoState videoState) {
 	if (!videoDecoder)
 		return;
 
+	videoDecoder->start();
+
 	byte *scaleBuffer = 0;
 	byte bytesPerPixel = videoDecoder->getPixelFormat().bytesPerPixel;
 	uint16 width = videoDecoder->getWidth();
@@ -162,9 +164,8 @@ reg_t kShowMovie(EngineState *s, int argc, reg_t *argv) {
 		} else {
 			// DOS SEQ
 			// SEQ's are called with no subops, just the string and delay
-			SeqDecoder *seqDecoder = new SeqDecoder();
-			seqDecoder->setFrameDelay(argv[1].toUint16()); // Time between frames in ticks
-			videoDecoder = seqDecoder;
+			// Time is specified as ticks
+			videoDecoder = new SEQDecoder(argv[1].toUint16());
 
 			if (!videoDecoder->loadFile(filename)) {
 				warning("Failed to open movie file %s", filename.c_str());
@@ -190,7 +191,7 @@ reg_t kShowMovie(EngineState *s, int argc, reg_t *argv) {
 		switch (argv[0].toUint16()) {
 		case 0: {
 			Common::String filename = s->_segMan->getString(argv[1]);
-			videoDecoder = new Video::AviDecoder(g_system->getMixer());
+			videoDecoder = new Video::AVIDecoder();
 
 			if (filename.equalsIgnoreCase("gk2a.avi")) {
 				// HACK: Switch to 16bpp graphics for Indeo3.
@@ -252,6 +253,7 @@ reg_t kRobot(EngineState *s, int argc, reg_t *argv) {
 		int16 y = argv[5].toUint16();
 		warning("kRobot(init), id %d, obj %04x:%04x, flag %d, x=%d, y=%d", id, PRINT_REG(obj), flag, x, y);
 		g_sci->_robotDecoder->load(id);
+		g_sci->_robotDecoder->start();
 		g_sci->_robotDecoder->setPos(x, y);
 		}
 		break;
@@ -267,13 +269,13 @@ reg_t kRobot(EngineState *s, int argc, reg_t *argv) {
 		warning("kRobot(%d)", subop);
 		break;
 	case 8: // sync
-		//if (false) {	// debug: automatically skip all robot videos
-		if ((uint32)g_sci->_robotDecoder->getCurFrame() !=  g_sci->_robotDecoder->getFrameCount() - 1) {
-			writeSelector(s->_segMan, argv[1], SELECTOR(signal), NULL_REG);
-		} else {
+		//if (true) {	// debug: automatically skip all robot videos
+		if (g_sci->_robotDecoder->endOfVideo()) {
 			g_sci->_robotDecoder->close();
 			// Signal the engine scripts that the video is done
 			writeSelector(s->_segMan, argv[1], SELECTOR(signal), SIGNAL_REG);
+		} else {
+			writeSelector(s->_segMan, argv[1], SELECTOR(signal), NULL_REG);	
 		}
 		break;
 	default:
@@ -348,7 +350,7 @@ reg_t kPlayVMD(EngineState *s, int argc, reg_t *argv) {
 		break;
 	}
 	case 6:	// Play
-		videoDecoder = new Video::VMDDecoder(g_system->getMixer());
+		videoDecoder = new Video::AdvancedVMDDecoder();
 
 		if (s->_videoState.fileName.empty()) {
 			// Happens in Lighthouse
@@ -406,7 +408,7 @@ reg_t kPlayDuck(EngineState *s, int argc, reg_t *argv) {
 		s->_videoState.reset();
 		s->_videoState.fileName = Common::String::format("%d.duk", argv[1].toUint16());
 
-		videoDecoder = new Video::AviDecoder(g_system->getMixer());
+		videoDecoder = new Video::AVIDecoder();
 
 		if (!videoDecoder->loadFile(s->_videoState.fileName)) {
 			warning("Could not open Duck %s", s->_videoState.fileName.c_str());
diff --git a/engines/sci/graphics/frameout.cpp b/engines/sci/graphics/frameout.cpp
index 6628247127..968014c032 100644
--- a/engines/sci/graphics/frameout.cpp
+++ b/engines/sci/graphics/frameout.cpp
@@ -28,6 +28,7 @@
 #include "common/system.h"
 #include "common/textconsole.h"
 #include "engines/engine.h"
+#include "graphics/palette.h"
 #include "graphics/surface.h"
 
 #include "sci/sci.h"
@@ -488,7 +489,7 @@ void GfxFrameout::showVideo() {
 	uint16 y = videoDecoder->getPos().y;
 
 	if (videoDecoder->hasDirtyPalette())
-		videoDecoder->setSystemPalette();
+		g_system->getPaletteManager()->setPalette(videoDecoder->getPalette(), 0, 256);
 
 	while (!g_engine->shouldQuit() && !videoDecoder->endOfVideo() && !skipVideo) {
 		if (videoDecoder->needsUpdate()) {
@@ -497,7 +498,7 @@ void GfxFrameout::showVideo() {
 				g_system->copyRectToScreen(frame->pixels, frame->pitch, x, y, frame->w, frame->h);
 
 				if (videoDecoder->hasDirtyPalette())
-					videoDecoder->setSystemPalette();
+					g_system->getPaletteManager()->setPalette(videoDecoder->getPalette(), 0, 256);
 
 				g_system->updateScreen();
 			}
diff --git a/engines/sci/sci.cpp b/engines/sci/sci.cpp
index d43a9d06fc..42ae00b525 100644
--- a/engines/sci/sci.cpp
+++ b/engines/sci/sci.cpp
@@ -632,7 +632,7 @@ void SciEngine::initGraphics() {
 		_gfxPaint = _gfxPaint32;
 		_gfxText32 = new GfxText32(_gamestate->_segMan, _gfxCache, _gfxScreen);
 		_gfxControls32 = new GfxControls32(_gamestate->_segMan, _gfxCache, _gfxScreen, _gfxText32);
-		_robotDecoder = new RobotDecoder(g_system->getMixer(), getPlatform() == Common::kPlatformMacintosh);
+		_robotDecoder = new RobotDecoder(getPlatform() == Common::kPlatformMacintosh);
 		_gfxFrameout = new GfxFrameout(_gamestate->_segMan, _resMan, _gfxCoordAdjuster, _gfxCache, _gfxScreen, _gfxPalette, _gfxPaint32);
 	} else {
 #endif
diff --git a/engines/sci/video/robot_decoder.cpp b/engines/sci/video/robot_decoder.cpp
index ebcfac6054..608c77136f 100644
--- a/engines/sci/video/robot_decoder.cpp
+++ b/engines/sci/video/robot_decoder.cpp
@@ -22,11 +22,13 @@
 
 #include "common/archive.h"
 #include "common/stream.h"
+#include "common/substream.h"
 #include "common/system.h"
 #include "common/textconsole.h"
 #include "common/util.h"
 
 #include "graphics/surface.h"
+#include "audio/audiostream.h"
 #include "audio/decoders/raw.h"
 
 #include "sci/resource.h"
@@ -63,57 +65,26 @@ namespace Sci {
 // our graphics engine, it looks just like a part of the room. A RBT can move
 // around the screen and go behind other objects. (...)
 
-#ifdef ENABLE_SCI32
-
-enum robotPalTypes {
+enum RobotPalTypes {
 	kRobotPalVariable = 0,
 	kRobotPalConstant = 1
 };
 
-RobotDecoder::RobotDecoder(Audio::Mixer *mixer, bool isBigEndian) {
-	_surface = 0;
-	_width = 0;
-	_height = 0;
+RobotDecoder::RobotDecoder(bool isBigEndian) {
 	_fileStream = 0;
-	_audioStream = 0;
-	_dirtyPalette = false;
 	_pos = Common::Point(0, 0);
-	_mixer = mixer;
 	_isBigEndian = isBigEndian;
+	_frameTotalSize = 0;
 }
 
 RobotDecoder::~RobotDecoder() {
 	close();
 }
 
-bool RobotDecoder::load(GuiResourceId id) {
-	// TODO: RAMA's robot 1003 cannot be played (shown at the menu screen) - 
-	// its drawn at odd coordinates. SV can't play it either (along with some
-	// others), so it must be some new functionality added in RAMA's robot
-	// videos. Skip it for now.
-	if (g_sci->getGameId() == GID_RAMA && id == 1003)
-		return false;
-	
-	// TODO: The robot video in the Lighthouse demo gets stuck
-	if (g_sci->getGameId() == GID_LIGHTHOUSE && id == 16)
-		return false;
-
-	Common::String fileName = Common::String::format("%d.rbt", id);
-	Common::SeekableReadStream *stream = SearchMan.createReadStreamForMember(fileName);
-
-	if (!stream) {
-		warning("Unable to open robot file %s", fileName.c_str());
-		return false;
-	}
-
-	return loadStream(stream);
-}
-
 bool RobotDecoder::loadStream(Common::SeekableReadStream *stream) {
 	close();
 
 	_fileStream = new Common::SeekableSubReadStreamEndian(stream, 0, stream->size(), _isBigEndian, DisposeAfterUse::YES);
-	_surface = new Graphics::Surface();
 
 	readHeaderChunk();
 
@@ -125,131 +96,60 @@ bool RobotDecoder::loadStream(Common::SeekableReadStream *stream) {
 	if (_header.version < 4 || _header.version > 6)
 		error("Unknown robot version: %d", _header.version);
 
-	if (_header.hasSound) {
-		_audioStream = Audio::makeQueuingAudioStream(11025, false);
-		_mixer->playStream(Audio::Mixer::kMusicSoundType, &_audioHandle, _audioStream, -1, getVolume(), getBalance());
-	}
+	RobotVideoTrack *videoTrack = new RobotVideoTrack(_header.frameCount);
+	addTrack(videoTrack);
 
-	readPaletteChunk(_header.paletteDataSize);
-	readFrameSizesChunk();
-	calculateVideoDimensions();
-	_surface->create(_width, _height, Graphics::PixelFormat::createFormatCLUT8());
+	if (_header.hasSound)
+		addTrack(new RobotAudioTrack());
 
+	videoTrack->readPaletteChunk(_fileStream, _header.paletteDataSize);
+	readFrameSizesChunk();
+	videoTrack->calculateVideoDimensions(_fileStream, _frameTotalSize);
 	return true;
 }
 
-void RobotDecoder::readHeaderChunk() {
-	// Header (60 bytes)
-	_fileStream->skip(6);
-	_header.version = _fileStream->readUint16();
-	_header.audioChunkSize = _fileStream->readUint16();
-	_header.audioSilenceSize = _fileStream->readUint16();
-	_fileStream->skip(2);
-	_header.frameCount = _fileStream->readUint16();
-	_header.paletteDataSize = _fileStream->readUint16();
-	_header.unkChunkDataSize = _fileStream->readUint16();
-	_fileStream->skip(5);
-	_header.hasSound = _fileStream->readByte();
-	_fileStream->skip(34);
-
-	// Some videos (e.g. robot 1305 in Phantasmagoria and
-	// robot 184 in Lighthouse) have an unknown chunk before
-	// the palette chunk (probably used for sound preloading).
-	// Skip it here.
-	if (_header.unkChunkDataSize)
-		_fileStream->skip(_header.unkChunkDataSize);
-}
-
-void RobotDecoder::readPaletteChunk(uint16 chunkSize) {
-	byte *paletteData = new byte[chunkSize];
-	_fileStream->read(paletteData, chunkSize);
-
-	// SCI1.1 palette
-	byte palFormat = paletteData[32];
-	uint16 palColorStart = paletteData[25];
-	uint16 palColorCount = READ_SCI11ENDIAN_UINT16(paletteData + 29);
+bool RobotDecoder::load(GuiResourceId id) {
+	// TODO: RAMA's robot 1003 cannot be played (shown at the menu screen) - 
+	// its drawn at odd coordinates. SV can't play it either (along with some
+	// others), so it must be some new functionality added in RAMA's robot
+	// videos. Skip it for now.
+	if (g_sci->getGameId() == GID_RAMA && id == 1003)
+		return false;
+	
+	// TODO: The robot video in the Lighthouse demo gets stuck
+	if (g_sci->getGameId() == GID_LIGHTHOUSE && id == 16)
+		return false;
 
-	int palOffset = 37;
-	memset(_palette, 0, 256 * 3);
+	Common::String fileName = Common::String::format("%d.rbt", id);
+	Common::SeekableReadStream *stream = SearchMan.createReadStreamForMember(fileName);
 
-	for (uint16 colorNo = palColorStart; colorNo < palColorStart + palColorCount; colorNo++) {
-		if (palFormat == kRobotPalVariable)
-			palOffset++;
-		_palette[colorNo * 3 + 0] = paletteData[palOffset++];
-		_palette[colorNo * 3 + 1] = paletteData[palOffset++];
-		_palette[colorNo * 3 + 2] = paletteData[palOffset++];
+	if (!stream) {
+		warning("Unable to open robot file %s", fileName.c_str());
+		return false;
 	}
 
-	_dirtyPalette = true;
-	delete[] paletteData;
+	return loadStream(stream);
 }
 
+void RobotDecoder::close() {
+	VideoDecoder::close();
 
-void RobotDecoder::readFrameSizesChunk() {
-	// The robot video file contains 2 tables, with one entry for each frame:
-	// - A table containing the size of the image in each video frame
-	// - A table containing the total size of each video frame.
-	// In v5 robots, the tables contain 16-bit integers, whereas in v6 robots,
-	// they contain 32-bit integers.
-
-	_frameTotalSize = new uint32[_header.frameCount];
-
-	// TODO: The table reading code can probably be removed once the
-	// audio chunk size is figured out (check the TODO inside processNextFrame())
-#if 0
-	// We don't need any of the two tables to play the video, so we ignore
-	// both of them.
-	uint16 wordSize = _header.version == 6 ? 4 : 2;
-	_fileStream->skip(_header.frameCount * wordSize * 2);
-#else
-	switch (_header.version) {
-	case 4:
-	case 5:		// sizes are 16-bit integers
-		// Skip table with frame image sizes, as we don't need it
-		_fileStream->skip(_header.frameCount * 2);
-		for (int i = 0; i < _header.frameCount; ++i)
-			_frameTotalSize[i] = _fileStream->readUint16();
-		break;
-	case 6:		// sizes are 32-bit integers
-		// Skip table with frame image sizes, as we don't need it
-		_fileStream->skip(_header.frameCount * 4);
-		for (int i = 0; i < _header.frameCount; ++i)
-			_frameTotalSize[i] = _fileStream->readUint32();
-		break;
-	default:
-		error("Can't yet handle index table for robot version %d", _header.version);
-	}
-#endif
-
-	// 2 more unknown tables
-	_fileStream->skip(1024 + 512);
+	delete _fileStream;
+	_fileStream = 0;
 
-	// Pad to nearest 2 kilobytes
-	uint32 curPos = _fileStream->pos();
-	if (curPos & 0x7ff)
-		_fileStream->seek((curPos & ~0x7ff) + 2048);
+	delete[] _frameTotalSize;
+	_frameTotalSize = 0;
 }
 
-void RobotDecoder::calculateVideoDimensions() {
-	// This is an O(n) operation, as each frame has a different size.
-	// We need to know the actual frame size to have a constant video size.
-	uint32 pos = _fileStream->pos();
-
-	for (uint32 curFrame = 0; curFrame < _header.frameCount; curFrame++) {
-		_fileStream->skip(4);
-		uint16 frameWidth = _fileStream->readUint16();
-		uint16 frameHeight = _fileStream->readUint16();
-		if (frameWidth > _width)
-			_width = frameWidth;
-		if (frameHeight > _height)
-			_height = frameHeight;
-		_fileStream->skip(_frameTotalSize[curFrame] - 8);
-	}
+void RobotDecoder::readNextPacket() {
+	// Get our track
+	RobotVideoTrack *videoTrack = (RobotVideoTrack *)getTrack(0);
+	videoTrack->increaseCurFrame();
+	Graphics::Surface *surface = videoTrack->getSurface();
 
-	_fileStream->seek(pos);
-}
+	if (videoTrack->endOfTrack())
+		return;
 
-const Graphics::Surface *RobotDecoder::decodeNextFrame() {
 	// Read frame image header (24 bytes)
 	_fileStream->skip(3);
 	byte frameScale = _fileStream->readByte();
@@ -258,23 +158,28 @@ const Graphics::Surface *RobotDecoder::decodeNextFrame() {
 	_fileStream->skip(4); // unknown, almost always 0
 	uint16 frameX = _fileStream->readUint16();
 	uint16 frameY = _fileStream->readUint16();
+
 	// TODO: In v4 robot files, frameX and frameY have a different meaning.
 	// Set them both to 0 for v4 for now, so that robots in PQ:SWAT show up
 	// correctly.
 	if (_header.version == 4)
 		frameX = frameY = 0;
+
 	uint16 compressedSize = _fileStream->readUint16();
 	uint16 frameFragments = _fileStream->readUint16();
 	_fileStream->skip(4); // unknown
 	uint32 decompressedSize = frameWidth * frameHeight * frameScale / 100;
+
 	// FIXME: A frame's height + position can go off limits... why? With the
 	// following, we cut the contents to fit the frame
-	uint16 scaledHeight = CLIP<uint16>(decompressedSize / frameWidth, 0, _height - frameY);
+	uint16 scaledHeight = CLIP<uint16>(decompressedSize / frameWidth, 0, surface->h - frameY);
+
 	// FIXME: Same goes for the frame's width + position. In this case, we
 	// modify the position to fit the contents on screen.
-	if (frameWidth + frameX > _width)
-		frameX = _width - frameWidth;
-	assert (frameWidth + frameX <= _width && scaledHeight + frameY <= _height);
+	if (frameWidth + frameX > surface->w)
+		frameX = surface->w - frameWidth;
+
+	assert(frameWidth + frameX <= surface->w && scaledHeight + frameY <= surface->h);
 
 	DecompressorLZS lzs;
 	byte *decompressedFrame = new byte[decompressedSize];
@@ -305,24 +210,23 @@ const Graphics::Surface *RobotDecoder::decodeNextFrame() {
 
 	// Copy over the decompressed frame
 	byte *inFrame = decompressedFrame;
-	byte *outFrame = (byte *)_surface->pixels;
+	byte *outFrame = (byte *)surface->pixels;
 
 	// Black out the surface
-	memset(outFrame, 0, _width * _height);
+	memset(outFrame, 0, surface->w * surface->h);
 
 	// Move to the correct y coordinate
-	outFrame += _width * frameY;
+	outFrame += surface->w * frameY;
 
 	for (uint16 y = 0; y < scaledHeight; y++) {
 		memcpy(outFrame + frameX, inFrame, frameWidth);
 		inFrame += frameWidth;
-		outFrame += _width;
+		outFrame += surface->w;
 	}
 
 	delete[] decompressedFrame;
 
-	// +1 because we start with frame number -1
-	uint32 audioChunkSize = _frameTotalSize[_curFrame + 1] - (24 + compressedSize);
+	uint32 audioChunkSize = _frameTotalSize[videoTrack->getCurFrame()] - (24 + compressedSize);
 
 // TODO: The audio chunk size below is usually correct, but there are some
 // exceptions (e.g. robot 4902 in Phantasmagoria, towards its end)
@@ -337,51 +241,166 @@ const Graphics::Surface *RobotDecoder::decodeNextFrame() {
 	// Queue the next audio frame
 	// FIXME: For some reason, there are audio hiccups/gaps
 	if (_header.hasSound) {
-		_fileStream->skip(8);	// header
-		_audioStream->queueBuffer(g_sci->_audio->getDecodedRobotAudioFrame(_fileStream, audioChunkSize - 8),
-									(audioChunkSize - 8) * 2, DisposeAfterUse::NO,
-									Audio::FLAG_16BITS | Audio::FLAG_LITTLE_ENDIAN);
+		RobotAudioTrack *audioTrack = (RobotAudioTrack *)getTrack(1);
+		_fileStream->skip(8); // header
+		audioChunkSize -= 8;
+		audioTrack->queueBuffer(g_sci->_audio->getDecodedRobotAudioFrame(_fileStream, audioChunkSize), audioChunkSize * 2);
 	} else {
 		_fileStream->skip(audioChunkSize);
-	}
-
-	if (_curFrame == -1)
-		_startTime = g_system->getMillis();
+	}	
+}
 
-	_curFrame++;
+void RobotDecoder::readHeaderChunk() {
+	// Header (60 bytes)
+	_fileStream->skip(6);
+	_header.version = _fileStream->readUint16();
+	_header.audioChunkSize = _fileStream->readUint16();
+	_header.audioSilenceSize = _fileStream->readUint16();
+	_fileStream->skip(2);
+	_header.frameCount = _fileStream->readUint16();
+	_header.paletteDataSize = _fileStream->readUint16();
+	_header.unkChunkDataSize = _fileStream->readUint16();
+	_fileStream->skip(5);
+	_header.hasSound = _fileStream->readByte();
+	_fileStream->skip(34);
 
-	return _surface;
+	// Some videos (e.g. robot 1305 in Phantasmagoria and
+	// robot 184 in Lighthouse) have an unknown chunk before
+	// the palette chunk (probably used for sound preloading).
+	// Skip it here.
+	if (_header.unkChunkDataSize)
+		_fileStream->skip(_header.unkChunkDataSize);
 }
 
-void RobotDecoder::close() {
-	if (!_fileStream)
-		return;
+void RobotDecoder::readFrameSizesChunk() {
+	// The robot video file contains 2 tables, with one entry for each frame:
+	// - A table containing the size of the image in each video frame
+	// - A table containing the total size of each video frame.
+	// In v5 robots, the tables contain 16-bit integers, whereas in v6 robots,
+	// they contain 32-bit integers.
 
-	delete _fileStream;
-	_fileStream = 0;
+	_frameTotalSize = new uint32[_header.frameCount];
 
+	// TODO: The table reading code can probably be removed once the
+	// audio chunk size is figured out (check the TODO inside processNextFrame())
+#if 0
+	// We don't need any of the two tables to play the video, so we ignore
+	// both of them.
+	uint16 wordSize = _header.version == 6 ? 4 : 2;
+	_fileStream->skip(_header.frameCount * wordSize * 2);
+#else
+	switch (_header.version) {
+	case 4:
+	case 5:		// sizes are 16-bit integers
+		// Skip table with frame image sizes, as we don't need it
+		_fileStream->skip(_header.frameCount * 2);
+		for (int i = 0; i < _header.frameCount; ++i)
+			_frameTotalSize[i] = _fileStream->readUint16();
+		break;
+	case 6:		// sizes are 32-bit integers
+		// Skip table with frame image sizes, as we don't need it
+		_fileStream->skip(_header.frameCount * 4);
+		for (int i = 0; i < _header.frameCount; ++i)
+			_frameTotalSize[i] = _fileStream->readUint32();
+		break;
+	default:
+		error("Can't yet handle index table for robot version %d", _header.version);
+	}
+#endif
+
+	// 2 more unknown tables
+	_fileStream->skip(1024 + 512);
+
+	// Pad to nearest 2 kilobytes
+	uint32 curPos = _fileStream->pos();
+	if (curPos & 0x7ff)
+		_fileStream->seek((curPos & ~0x7ff) + 2048);
+}
+
+RobotDecoder::RobotVideoTrack::RobotVideoTrack(int frameCount) : _frameCount(frameCount) {
+	_surface = new Graphics::Surface();
+	_curFrame = -1;
+	_dirtyPalette = false;
+}
+
+RobotDecoder::RobotVideoTrack::~RobotVideoTrack() {
 	_surface->free();
 	delete _surface;
-	_surface = 0;
+}
 
-	if (_header.hasSound) {
-		_mixer->stopHandle(_audioHandle);
-		//delete _audioStream; _audioStream = 0;
+uint16 RobotDecoder::RobotVideoTrack::getWidth() const {
+	return _surface->w;
+}
+
+uint16 RobotDecoder::RobotVideoTrack::getHeight() const {
+	return _surface->h;
+}
+
+Graphics::PixelFormat RobotDecoder::RobotVideoTrack::getPixelFormat() const {
+	return _surface->format;
+}
+
+void RobotDecoder::RobotVideoTrack::readPaletteChunk(Common::SeekableSubReadStreamEndian *stream, uint16 chunkSize) {
+	byte *paletteData = new byte[chunkSize];
+	stream->read(paletteData, chunkSize);
+
+	// SCI1.1 palette
+	byte palFormat = paletteData[32];
+	uint16 palColorStart = paletteData[25];
+	uint16 palColorCount = READ_SCI11ENDIAN_UINT16(paletteData + 29);
+
+	int palOffset = 37;
+	memset(_palette, 0, 256 * 3);
+
+	for (uint16 colorNo = palColorStart; colorNo < palColorStart + palColorCount; colorNo++) {
+		if (palFormat == kRobotPalVariable)
+			palOffset++;
+		_palette[colorNo * 3 + 0] = paletteData[palOffset++];
+		_palette[colorNo * 3 + 1] = paletteData[palOffset++];
+		_palette[colorNo * 3 + 2] = paletteData[palOffset++];
 	}
 
-	reset();
+	_dirtyPalette = true;
+	delete[] paletteData;
 }
 
-void RobotDecoder::updateVolume() {
-	if (g_system->getMixer()->isSoundHandleActive(_audioHandle))
-		g_system->getMixer()->setChannelVolume(_audioHandle, getVolume());
+void RobotDecoder::RobotVideoTrack::calculateVideoDimensions(Common::SeekableSubReadStreamEndian *stream, uint32 *frameSizes) {
+	// This is an O(n) operation, as each frame has a different size.
+	// We need to know the actual frame size to have a constant video size.
+	uint32 pos = stream->pos();
+
+	uint16 width = 0, height = 0;
+
+	for (int curFrame = 0; curFrame < _frameCount; curFrame++) {
+		stream->skip(4);
+		uint16 frameWidth = stream->readUint16();
+		uint16 frameHeight = stream->readUint16();
+		if (frameWidth > width)
+			width = frameWidth;
+		if (frameHeight > height)
+			height = frameHeight;
+		stream->skip(frameSizes[curFrame] - 8);
+	}
+
+	stream->seek(pos);
+
+	_surface->create(width, height, Graphics::PixelFormat::createFormatCLUT8());
 }
 
-void RobotDecoder::updateBalance() {
-	if (g_system->getMixer()->isSoundHandleActive(_audioHandle))
-		g_system->getMixer()->setChannelBalance(_audioHandle, getBalance());
+RobotDecoder::RobotAudioTrack::RobotAudioTrack() {
+	_audioStream = Audio::makeQueuingAudioStream(11025, false);
 }
 
-#endif
+RobotDecoder::RobotAudioTrack::~RobotAudioTrack() {
+	delete _audioStream;
+}
+
+void RobotDecoder::RobotAudioTrack::queueBuffer(byte *buffer, int size) {
+	_audioStream->queueBuffer(buffer, size, DisposeAfterUse::YES, Audio::FLAG_16BITS | Audio::FLAG_LITTLE_ENDIAN);
+}
+
+Audio::AudioStream *RobotDecoder::RobotAudioTrack::getAudioStream() const {
+	return _audioStream;
+}
 
 } // End of namespace Sci
diff --git a/engines/sci/video/robot_decoder.h b/engines/sci/video/robot_decoder.h
index e9cefe7d91..ebc3262939 100644
--- a/engines/sci/video/robot_decoder.h
+++ b/engines/sci/video/robot_decoder.h
@@ -25,84 +25,103 @@
 
 #include "common/rational.h"
 #include "common/rect.h"
-#include "common/stream.h"
-#include "common/substream.h"
-#include "audio/audiostream.h"
-#include "audio/mixer.h"
-#include "graphics/pixelformat.h"
 #include "video/video_decoder.h"
 
-namespace Sci {
+namespace Audio {
+class QueuingAudioStream;
+}
 
-#ifdef ENABLE_SCI32
-
-struct RobotHeader {
-	// 6 bytes, identifier bytes
-	uint16 version;
-	uint16 audioChunkSize;
-	uint16 audioSilenceSize;
-	// 2 bytes, unknown
-	uint16 frameCount;
-	uint16 paletteDataSize;
-	uint16 unkChunkDataSize;
-	// 5 bytes, unknown
-	byte hasSound;
-	// 34 bytes, unknown
-};
+namespace Common {
+class SeekableSubReadStreamEndian;
+}
+
+namespace Sci {
 
-class RobotDecoder : public Video::FixedRateVideoDecoder {
+class RobotDecoder : public Video::VideoDecoder {
 public:
-	RobotDecoder(Audio::Mixer *mixer, bool isBigEndian);
+	RobotDecoder(bool isBigEndian);
 	virtual ~RobotDecoder();
 
 	bool loadStream(Common::SeekableReadStream *stream);
 	bool load(GuiResourceId id);
 	void close();
-
-	bool isVideoLoaded() const { return _fileStream != 0; }
-	uint16 getWidth() const { return _width; }
-	uint16 getHeight() const { return _height; }
-	uint32 getFrameCount() const { return _header.frameCount; }
-	const Graphics::Surface *decodeNextFrame();
-	Graphics::PixelFormat getPixelFormat() const { return Graphics::PixelFormat::createFormatCLUT8(); }
-	const byte *getPalette() { _dirtyPalette = false; return _palette; }
-	bool hasDirtyPalette() const { return _dirtyPalette; }
+	
 	void setPos(uint16 x, uint16 y) { _pos = Common::Point(x, y); }
 	Common::Point getPos() const { return _pos; }
 
 protected:
-	// VideoDecoder API
-	void updateVolume();
-	void updateBalance();
-
-	// FixedRateVideoDecoder API
-	Common::Rational getFrameRate() const { return Common::Rational(60, 10); }
-
+	void readNextPacket();
+	
 private:
+	class RobotVideoTrack : public FixedRateVideoTrack {
+	public:
+		RobotVideoTrack(int frameCount);
+		~RobotVideoTrack();
+
+		uint16 getWidth() const;
+		uint16 getHeight() const;
+		Graphics::PixelFormat getPixelFormat() const;
+		int getCurFrame() const { return _curFrame; }
+		int getFrameCount() const { return _frameCount; }
+		const Graphics::Surface *decodeNextFrame() { return _surface; }
+		const byte *getPalette() const { _dirtyPalette = false; return _palette; }
+		bool hasDirtyPalette() const { return _dirtyPalette; }
+
+		void readPaletteChunk(Common::SeekableSubReadStreamEndian *stream, uint16 chunkSize);
+		void calculateVideoDimensions(Common::SeekableSubReadStreamEndian *stream, uint32 *frameSizes);
+		Graphics::Surface *getSurface() { return _surface; }
+		void increaseCurFrame() { _curFrame++; }
+
+	protected:
+		Common::Rational getFrameRate() const { return Common::Rational(60, 10); }
+
+	private:
+		int _frameCount;
+		int _curFrame;
+		byte _palette[256 * 3];
+		mutable bool _dirtyPalette;
+		Graphics::Surface *_surface;
+	};
+
+	class RobotAudioTrack : public AudioTrack {
+	public:
+		RobotAudioTrack();
+		~RobotAudioTrack();
+
+		Audio::Mixer::SoundType getSoundType() const { return Audio::Mixer::kMusicSoundType; }
+
+		void queueBuffer(byte *buffer, int size);
+
+	protected:
+		Audio::AudioStream *getAudioStream() const;
+
+	private:
+		Audio::QueuingAudioStream *_audioStream;
+	};
+
+	struct RobotHeader {
+		// 6 bytes, identifier bytes
+		uint16 version;
+		uint16 audioChunkSize;
+		uint16 audioSilenceSize;
+		// 2 bytes, unknown
+		uint16 frameCount;
+		uint16 paletteDataSize;
+		uint16 unkChunkDataSize;
+		// 5 bytes, unknown
+		byte hasSound;
+		// 34 bytes, unknown
+	} _header;
+
 	void readHeaderChunk();
-	void readPaletteChunk(uint16 chunkSize);
 	void readFrameSizesChunk();
-	void calculateVideoDimensions();
-
-	void freeData();
 
-	RobotHeader _header;
 	Common::Point _pos;
 	bool _isBigEndian;
+	uint32 *_frameTotalSize;
 
 	Common::SeekableSubReadStreamEndian *_fileStream;
-
-	uint16 _width;
-	uint16 _height;
-	uint32 *_frameTotalSize;
-	byte _palette[256 * 3];
-	bool _dirtyPalette;
-	Graphics::Surface *_surface;
-	Audio::QueuingAudioStream *_audioStream;
-	Audio::SoundHandle _audioHandle;
-	Audio::Mixer *_mixer;
 };
-#endif
 
 } // End of namespace Sci
 
diff --git a/engines/sci/video/seq_decoder.cpp b/engines/sci/video/seq_decoder.cpp
index abd64911a7..a7b6346eca 100644
--- a/engines/sci/video/seq_decoder.cpp
+++ b/engines/sci/video/seq_decoder.cpp
@@ -41,33 +41,44 @@ enum seqFrameTypes {
 	kSeqFrameDiff = 1
 };
 
-SeqDecoder::SeqDecoder() {
-	_fileStream = 0;
-	_surface = 0;
-	_dirtyPalette = false;
+SEQDecoder::SEQDecoder(uint frameDelay) : _frameDelay(frameDelay) {
 }
 
-SeqDecoder::~SeqDecoder() {
+SEQDecoder::~SEQDecoder() {
 	close();
 }
 
-bool SeqDecoder::loadStream(Common::SeekableReadStream *stream) {
+bool SEQDecoder::loadStream(Common::SeekableReadStream *stream) {
 	close();
 
+	addTrack(new SEQVideoTrack(stream, _frameDelay));
+
+	return true;
+}
+
+SEQDecoder::SEQVideoTrack::SEQVideoTrack(Common::SeekableReadStream *stream, uint frameDelay) {
+	assert(stream);
+	assert(frameDelay != 0);
 	_fileStream = stream;
+	_frameDelay = frameDelay;
+	_curFrame = -1;
+
 	_surface = new Graphics::Surface();
 	_surface->create(SEQ_SCREEN_WIDTH, SEQ_SCREEN_HEIGHT, Graphics::PixelFormat::createFormatCLUT8());
 
 	_frameCount = _fileStream->readUint16LE();
 
-	// Set palette
-	int paletteChunkSize = _fileStream->readUint32LE();
-	readPaletteChunk(paletteChunkSize);
+	// Set initial palette
+	readPaletteChunk(_fileStream->readUint32LE());
+}
 
-	return true;
+SEQDecoder::SEQVideoTrack::~SEQVideoTrack() {
+	delete _fileStream;
+	_surface->free();
+	delete _surface;
 }
 
-void SeqDecoder::readPaletteChunk(uint16 chunkSize) {
+void SEQDecoder::SEQVideoTrack::readPaletteChunk(uint16 chunkSize) {
 	byte *paletteData = new byte[chunkSize];
 	_fileStream->read(paletteData, chunkSize);
 
@@ -91,23 +102,7 @@ void SeqDecoder::readPaletteChunk(uint16 chunkSize) {
 	delete[] paletteData;
 }
 
-void SeqDecoder::close() {
-	if (!_fileStream)
-		return;
-
-	_frameDelay = 0;
-
-	delete _fileStream;
-	_fileStream = 0;
-
-	_surface->free();
-	delete _surface;
-	_surface = 0;
-
-	reset();
-}
-
-const Graphics::Surface *SeqDecoder::decodeNextFrame() {
+const Graphics::Surface *SEQDecoder::SEQVideoTrack::decodeNextFrame() {
 	int16 frameWidth = _fileStream->readUint16LE();
 	int16 frameHeight = _fileStream->readUint16LE();
 	int16 frameLeft = _fileStream->readUint16LE();
@@ -142,9 +137,6 @@ const Graphics::Surface *SeqDecoder::decodeNextFrame() {
 		delete[] buf;
 	}
 
-	if (_curFrame == -1)
-		_startTime = g_system->getMillis();
-
 	_curFrame++;
 	return _surface;
 }
@@ -159,7 +151,7 @@ const Graphics::Surface *SeqDecoder::decodeNextFrame() {
 	} \
 	memcpy(dest + writeRow * SEQ_SCREEN_WIDTH + writeCol, litData + litPos, n);
 
-bool SeqDecoder::decodeFrame(byte *rleData, int rleSize, byte *litData, int litSize, byte *dest, int left, int width, int height, int colorKey) {
+bool SEQDecoder::SEQVideoTrack::decodeFrame(byte *rleData, int rleSize, byte *litData, int litSize, byte *dest, int left, int width, int height, int colorKey) {
 	int writeRow = 0;
 	int writeCol = left;
 	int litPos = 0;
@@ -237,4 +229,9 @@ bool SeqDecoder::decodeFrame(byte *rleData, int rleSize, byte *litData, int litS
 	return true;
 }
 
+const byte *SEQDecoder::SEQVideoTrack::getPalette() const {
+	_dirtyPalette = false;
+	return _palette;
+}
+
 } // End of namespace Sci
diff --git a/engines/sci/video/seq_decoder.h b/engines/sci/video/seq_decoder.h
index 800a3c9024..890f349feb 100644
--- a/engines/sci/video/seq_decoder.h
+++ b/engines/sci/video/seq_decoder.h
@@ -40,44 +40,49 @@ namespace Sci {
 /**
  * Implementation of the Sierra SEQ decoder, used in KQ6 DOS floppy/CD and GK1 DOS
  */
-class SeqDecoder : public Video::FixedRateVideoDecoder {
+class SEQDecoder : public Video::VideoDecoder {
 public:
-	SeqDecoder();
-	virtual ~SeqDecoder();
+	SEQDecoder(uint frameDelay);
+	virtual ~SEQDecoder();
 
 	bool loadStream(Common::SeekableReadStream *stream);
-	void close();
-
-	void setFrameDelay(int frameDelay) { _frameDelay = frameDelay; }
-
-	bool isVideoLoaded() const { return _fileStream != 0; }
-	uint16 getWidth() const { return SEQ_SCREEN_WIDTH; }
-	uint16 getHeight() const { return SEQ_SCREEN_HEIGHT; }
-	uint32 getFrameCount() const { return _frameCount; }
-	const Graphics::Surface *decodeNextFrame();
-	Graphics::PixelFormat getPixelFormat() const { return Graphics::PixelFormat::createFormatCLUT8(); }
-	const byte *getPalette() { _dirtyPalette = false; return _palette; }
-	bool hasDirtyPalette() const { return _dirtyPalette; }
-
-protected:
-	Common::Rational getFrameRate() const { assert(_frameDelay); return Common::Rational(60, _frameDelay); }
 
 private:
-	enum {
-		SEQ_SCREEN_WIDTH = 320,
-		SEQ_SCREEN_HEIGHT = 200
+	class SEQVideoTrack : public FixedRateVideoTrack {
+	public:
+		SEQVideoTrack(Common::SeekableReadStream *stream, uint frameDelay);
+		~SEQVideoTrack();
+
+		uint16 getWidth() const { return SEQ_SCREEN_WIDTH; }
+		uint16 getHeight() const { return SEQ_SCREEN_HEIGHT; }
+		Graphics::PixelFormat getPixelFormat() const { return Graphics::PixelFormat::createFormatCLUT8(); }
+		int getCurFrame() const { return _curFrame; }
+		int getFrameCount() const { return _frameCount; }
+		const Graphics::Surface *decodeNextFrame();
+		const byte *getPalette() const;
+		bool hasDirtyPalette() const { return _dirtyPalette; }
+
+	protected:
+		Common::Rational getFrameRate() const { return Common::Rational(60, _frameDelay); }
+
+	private:
+		enum {
+			SEQ_SCREEN_WIDTH = 320,
+			SEQ_SCREEN_HEIGHT = 200
+		};
+
+		void readPaletteChunk(uint16 chunkSize);
+		bool decodeFrame(byte *rleData, int rleSize, byte *litData, int litSize, byte *dest, int left, int width, int height, int colorKey);
+
+		Common::SeekableReadStream *_fileStream;
+		int _curFrame, _frameCount;
+		byte _palette[256 * 3];
+		mutable bool _dirtyPalette;
+		Graphics::Surface *_surface;
+		uint _frameDelay;
 	};
 
-	void readPaletteChunk(uint16 chunkSize);
-	bool decodeFrame(byte *rleData, int rleSize, byte *litData, int litSize, byte *dest, int left, int width, int height, int colorKey);
-
-	uint16 _width, _height;
-	uint16 _frameDelay;
-	Common::SeekableReadStream *_fileStream;
-	byte _palette[256 * 3];
-	bool _dirtyPalette;
-	uint32 _frameCount;
-	Graphics::Surface *_surface;
+	uint _frameDelay;
 };
 
 } // End of namespace Sci