h265 video mp4 support

codec/h265parser/parser.go

@@ -2,6 +2,7 @@ package h265parser
 
 import (
 	"bytes"
+	"errors"
 	"fmt"
 
 	"github.com/deepch/vdk/av"
@@ -9,6 +10,32 @@ import (
 	"github.com/deepch/vdk/utils/bits/pio"
 )
 
+type SPSInfo struct {
+	ProfileIdc                       uint
+	LevelIdc                         uint
+	MbWidth                          uint
+	MbHeight                         uint
+	CropLeft                         uint
+	CropRight                        uint
+	CropTop                          uint
+	CropBottom                       uint
+	Width                            uint
+	Height                           uint
+	numTemporalLayers                uint
+	temporalIdNested                 uint
+	chromaFormat                     uint
+	PicWidthInLumaSamples            uint
+	PicHeightInLumaSamples           uint
+	bitDepthLumaMinus8               uint
+	bitDepthChromaMinus8             uint
+	generalProfileSpace              uint
+	generalTierFlag                  uint
+	generalProfileIDC                uint
+	generalProfileCompatibilityFlags uint32
+	generalConstraintIndicatorFlags  uint64
+	generalLevelIDC                  uint
+}
+
 const (
 	NAL_UNIT_CODED_SLICE_TRAIL_N    = 0
 	NAL_UNIT_CODED_SLICE_TRAIL_R    = 1
@@ -83,6 +110,11 @@ const (
 	MAX_SPS_COUNT  = 32
 )
 
+var (
+	ErrorH265IncorectUnitSize = errors.New("Invorect Unit Size")
+	ErrorH265IncorectUnitType = errors.New("Incorect Unit Type")
+)
+
 func IsDataNALU(b []byte) bool {
 	typ := b[0] & 0x1f
 	return typ >= 1 && typ <= 5
@@ -128,7 +160,6 @@ func SplitNALUs(b []byte) (nalus [][]byte, typ int) {
 			return nalus, NALU_AVCC
 		}
 	}
-	// is Annex B
 	if val3 == 1 || val4 == 1 {
 		_val3 := val3
 		_val4 := val4
@@ -175,212 +206,216 @@ func SplitNALUs(b []byte) (nalus [][]byte, typ int) {
 	return [][]byte{b}, NALU_RAW
 }
 
-type SPSInfo struct {
+func ParseSPS(sps []byte) (ctx SPSInfo, err error) {
-	ProfileIdc uint
+	if len(sps) < 2 {
-	LevelIdc   uint
+		err = ErrorH265IncorectUnitSize
+		return
+	}
+	rbsp := nal2rbsp(sps[2:])
+	br := &bits.GolombBitReader{R: bytes.NewReader(rbsp)}
+	if _, err = br.ReadBits(4); err != nil {
+		return
+	}
+	spsMaxSubLayersMinus1, err := br.ReadBits(3)
+	if err != nil {
+		return
+	}
 
-	MbWidth  uint
+	if spsMaxSubLayersMinus1+1 > ctx.numTemporalLayers {
-	MbHeight uint
+		ctx.numTemporalLayers = spsMaxSubLayersMinus1 + 1
+	}
+	if ctx.temporalIdNested, err = br.ReadBit(); err != nil {
+		return
+	}
+	if err = parsePTL(br, &ctx, spsMaxSubLayersMinus1); err != nil {
+		return
+	}
+	if _, err = br.ReadExponentialGolombCode(); err != nil {
+		return
+	}
+	var cf uint
+	if cf, err = br.ReadExponentialGolombCode(); err != nil {
+		return
+	}
+	ctx.chromaFormat = uint(cf)
+	if ctx.chromaFormat == 3 {
+		if _, err = br.ReadBit(); err != nil {
+			return
+		}
+	}
+	if ctx.PicWidthInLumaSamples, err = br.ReadExponentialGolombCode(); err != nil {
+		return
+	}
+	ctx.Width = uint(ctx.PicWidthInLumaSamples)
+	if ctx.PicHeightInLumaSamples, err = br.ReadExponentialGolombCode(); err != nil {
+		return
+	}
+	ctx.Height = uint(ctx.PicHeightInLumaSamples)
+	conformanceWindowFlag, err := br.ReadBit()
+	if err != nil {
+		return
+	}
+	if conformanceWindowFlag != 0 {
+		if _, err = br.ReadExponentialGolombCode(); err != nil {
+			return
+		}
+		if _, err = br.ReadExponentialGolombCode(); err != nil {
+			return
+		}
+		if _, err = br.ReadExponentialGolombCode(); err != nil {
+			return
+		}
+		if _, err = br.ReadExponentialGolombCode(); err != nil {
+			return
+		}
+	}
 
-	CropLeft   uint
+	var bdlm8 uint
-	CropRight  uint
+	if bdlm8, err = br.ReadExponentialGolombCode(); err != nil {
-	CropTop    uint
+		return
-	CropBottom uint
+	}
+	ctx.bitDepthChromaMinus8 = uint(bdlm8)
+	var bdcm8 uint
+	if bdcm8, err = br.ReadExponentialGolombCode(); err != nil {
+		return
+	}
+	ctx.bitDepthChromaMinus8 = uint(bdcm8)
 
-	Width  uint
+	_, err = br.ReadExponentialGolombCode()
-	Height uint
+	if err != nil {
+		return
+	}
+	spsSubLayerOrderingInfoPresentFlag, err := br.ReadBit()
+	if err != nil {
+		return
+	}
+	var i uint
+	if spsSubLayerOrderingInfoPresentFlag != 0 {
+		i = 0
+	} else {
+		i = spsMaxSubLayersMinus1
+	}
+	for ; i <= spsMaxSubLayersMinus1; i++ {
+		if _, err = br.ReadExponentialGolombCode(); err != nil {
+			return
+		}
+		if _, err = br.ReadExponentialGolombCode(); err != nil {
+			return
+		}
+		if _, err = br.ReadExponentialGolombCode(); err != nil {
+			return
+		}
+	}
+
+	if _, err = br.ReadExponentialGolombCode(); err != nil {
+		return
+	}
+	if _, err = br.ReadExponentialGolombCode(); err != nil {
+		return
+	}
+	if _, err = br.ReadExponentialGolombCode(); err != nil {
+		return
+	}
+	if _, err = br.ReadExponentialGolombCode(); err != nil {
+		return
+	}
+	if _, err = br.ReadExponentialGolombCode(); err != nil {
+		return
+	}
+	if _, err = br.ReadExponentialGolombCode(); err != nil {
+		return
+	}
+	return
 }
 
-func ParseSPS(data []byte) (self SPSInfo, err error) {
+func parsePTL(br *bits.GolombBitReader, ctx *SPSInfo, maxSubLayersMinus1 uint) error {
-	r := &bits.GolombBitReader{R: bytes.NewReader(data)}
+	var err error
-
+	var ptl SPSInfo
-	if _, err = r.ReadBits(8); err != nil {
+	if ptl.generalProfileSpace, err = br.ReadBits(2); err != nil {
-		return
+		return err
 	}
-
+	if ptl.generalTierFlag, err = br.ReadBit(); err != nil {
-	if self.ProfileIdc, err = r.ReadBits(8); err != nil {
+		return err
-		return
 	}
-
+	if ptl.generalProfileIDC, err = br.ReadBits(5); err != nil {
-	// constraint_set0_flag-constraint_set6_flag,reserved_zero_2bits
+		return err
-	if _, err = r.ReadBits(8); err != nil {
-		return
 	}
-
+	if ptl.generalProfileCompatibilityFlags, err = br.ReadBits32(32); err != nil {
-	// level_idc
+		return err
-	if self.LevelIdc, err = r.ReadBits(8); err != nil {
-		return
 	}
-
+	if ptl.generalConstraintIndicatorFlags, err = br.ReadBits64(48); err != nil {
-	// seq_parameter_set_id
+		return err
-	if _, err = r.ReadExponentialGolombCode(); err != nil {
-		return
 	}
-
+	if ptl.generalLevelIDC, err = br.ReadBits(8); err != nil {
-	if self.ProfileIdc == 100 || self.ProfileIdc == 110 ||
+		return err
-		self.ProfileIdc == 122 || self.ProfileIdc == 244 ||
-		self.ProfileIdc == 44 || self.ProfileIdc == 83 ||
-		self.ProfileIdc == 86 || self.ProfileIdc == 118 {
-
-		var chroma_format_idc uint
-		if chroma_format_idc, err = r.ReadExponentialGolombCode(); err != nil {
-			return
 	}
-
+	updatePTL(ctx, &ptl)
-		if chroma_format_idc == 3 {
+	if maxSubLayersMinus1 == 0 {
-			// residual_colour_transform_flag
+		return nil
-			if _, err = r.ReadBit(); err != nil {
+	}
-				return
+	subLayerProfilePresentFlag := make([]uint, maxSubLayersMinus1)
+	subLayerLevelPresentFlag := make([]uint, maxSubLayersMinus1)
+	for i := uint(0); i < maxSubLayersMinus1; i++ {
+		if subLayerProfilePresentFlag[i], err = br.ReadBit(); err != nil {
+			return err
+		}
+		if subLayerLevelPresentFlag[i], err = br.ReadBit(); err != nil {
+			return err
+		}
+	}
+	if maxSubLayersMinus1 > 0 {
+		for i := maxSubLayersMinus1; i < 8; i++ {
+			if _, err = br.ReadBits(2); err != nil {
+				return err
+			}
+		}
+	}
+	for i := uint(0); i < maxSubLayersMinus1; i++ {
+		if subLayerProfilePresentFlag[i] != 0 {
+			if _, err = br.ReadBits32(32); err != nil {
+				return err
+			}
+			if _, err = br.ReadBits32(32); err != nil {
+				return err
+			}
+			if _, err = br.ReadBits32(24); err != nil {
+				return err
 			}
 		}
 
-		// bit_depth_luma_minus8
+		if subLayerLevelPresentFlag[i] != 0 {
-		if _, err = r.ReadExponentialGolombCode(); err != nil {
+			if _, err = br.ReadBits(8); err != nil {
-			return
+				return err
 			}
-		// bit_depth_chroma_minus8
-		if _, err = r.ReadExponentialGolombCode(); err != nil {
-			return
 		}
-		// qpprime_y_zero_transform_bypass_flag
-		if _, err = r.ReadBit(); err != nil {
-			return
 	}
+	return nil
+}
 
-		var seq_scaling_matrix_present_flag uint
+func updatePTL(ctx, ptl *SPSInfo) {
-		if seq_scaling_matrix_present_flag, err = r.ReadBit(); err != nil {
+	ctx.generalProfileSpace = ptl.generalProfileSpace
-			return
-		}
 
-		if seq_scaling_matrix_present_flag != 0 {
+	if ptl.generalTierFlag > ctx.generalTierFlag {
-			for i := 0; i < 8; i++ {
+		ctx.generalLevelIDC = ptl.generalLevelIDC
-				var seq_scaling_list_present_flag uint
+
-				if seq_scaling_list_present_flag, err = r.ReadBit(); err != nil {
+		ctx.generalTierFlag = ptl.generalTierFlag
-					return
-				}
-				if seq_scaling_list_present_flag != 0 {
-					var sizeOfScalingList uint
-					if i < 6 {
-						sizeOfScalingList = 16
 	} else {
-						sizeOfScalingList = 64
+		if ptl.generalLevelIDC > ctx.generalLevelIDC {
-					}
+			ctx.generalLevelIDC = ptl.generalLevelIDC
-					lastScale := uint(8)
-					nextScale := uint(8)
-					for j := uint(0); j < sizeOfScalingList; j++ {
-						if nextScale != 0 {
-							var delta_scale uint
-							if delta_scale, err = r.ReadSE(); err != nil {
-								return
-							}
-							nextScale = (lastScale + delta_scale + 256) % 256
-						}
-						if nextScale != 0 {
-							lastScale = nextScale
-						}
-					}
-				}
-			}
 		}
 	}
 
-	// log2_max_frame_num_minus4
+	if ptl.generalProfileIDC > ctx.generalProfileIDC {
-	if _, err = r.ReadExponentialGolombCode(); err != nil {
+		ctx.generalProfileIDC = ptl.generalProfileIDC
-		return
 	}
 
-	var pic_order_cnt_type uint
+	ctx.generalProfileCompatibilityFlags &= ptl.generalProfileCompatibilityFlags
-	if pic_order_cnt_type, err = r.ReadExponentialGolombCode(); err != nil {
-		return
-	}
-	if pic_order_cnt_type == 0 {
-		// log2_max_pic_order_cnt_lsb_minus4
-		if _, err = r.ReadExponentialGolombCode(); err != nil {
-			return
-		}
-	} else if pic_order_cnt_type == 1 {
-		// delta_pic_order_always_zero_flag
-		if _, err = r.ReadBit(); err != nil {
-			return
-		}
-		// offset_for_non_ref_pic
-		if _, err = r.ReadSE(); err != nil {
-			return
-		}
-		// offset_for_top_to_bottom_field
-		if _, err = r.ReadSE(); err != nil {
-			return
-		}
-		var num_ref_frames_in_pic_order_cnt_cycle uint
-		if num_ref_frames_in_pic_order_cnt_cycle, err = r.ReadExponentialGolombCode(); err != nil {
-			return
-		}
-		for i := uint(0); i < num_ref_frames_in_pic_order_cnt_cycle; i++ {
-			if _, err = r.ReadSE(); err != nil {
-				return
-			}
-		}
-	}
 
-	// max_num_ref_frames
+	ctx.generalConstraintIndicatorFlags &= ptl.generalConstraintIndicatorFlags
-	if _, err = r.ReadExponentialGolombCode(); err != nil {
+}
-		return
-	}
 
-	// gaps_in_frame_num_value_allowed_flag
+func nal2rbsp(nal []byte) []byte {
-	if _, err = r.ReadBit(); err != nil {
+	return bytes.Replace(nal, []byte{0x0, 0x0, 0x3}, []byte{0x0, 0x0}, -1)
-		return
-	}
-
-	if self.MbWidth, err = r.ReadExponentialGolombCode(); err != nil {
-		return
-	}
-	self.MbWidth++
-
-	if self.MbHeight, err = r.ReadExponentialGolombCode(); err != nil {
-		return
-	}
-	self.MbHeight++
-
-	var frame_mbs_only_flag uint
-	if frame_mbs_only_flag, err = r.ReadBit(); err != nil {
-		return
-	}
-	if frame_mbs_only_flag == 0 {
-		// mb_adaptive_frame_field_flag
-		if _, err = r.ReadBit(); err != nil {
-			return
-		}
-	}
-
-	// direct_8x8_inference_flag
-	if _, err = r.ReadBit(); err != nil {
-		return
-	}
-
-	var frame_cropping_flag uint
-	if frame_cropping_flag, err = r.ReadBit(); err != nil {
-		return
-	}
-	if frame_cropping_flag != 0 {
-		if self.CropLeft, err = r.ReadExponentialGolombCode(); err != nil {
-			return
-		}
-		if self.CropRight, err = r.ReadExponentialGolombCode(); err != nil {
-			return
-		}
-		if self.CropTop, err = r.ReadExponentialGolombCode(); err != nil {
-			return
-		}
-		if self.CropBottom, err = r.ReadExponentialGolombCode(); err != nil {
-			return
-		}
-	}
-
-	self.Width = (self.MbWidth * 16) - self.CropLeft*2 - self.CropRight*2
-	self.Height = ((2 - frame_mbs_only_flag) * self.MbHeight * 16) - self.CropTop*2 - self.CropBottom*2
-
-	return
 }
 
 type CodecData struct {
@@ -443,23 +478,20 @@ func NewCodecDataFromAVCDecoderConfRecord(record []byte) (self CodecData, err er
 
 func NewCodecDataFromVPSAndSPSAndPPS(vps, sps, pps []byte) (self CodecData, err error) {
 	recordinfo := AVCDecoderConfRecord{}
-	recordinfo.AVCProfileIndication = sps[1]
+	recordinfo.AVCProfileIndication = sps[3]
-	recordinfo.ProfileCompatibility = sps[2]
+	recordinfo.ProfileCompatibility = sps[4]
-	recordinfo.AVCLevelIndication = sps[3]
+	recordinfo.AVCLevelIndication = sps[5]
 	recordinfo.SPS = [][]byte{sps}
 	recordinfo.PPS = [][]byte{pps}
 	recordinfo.VPS = [][]byte{vps}
 	recordinfo.LengthSizeMinusOne = 3
-
-	buf := make([]byte, recordinfo.Len())
-	recordinfo.Marshal(buf)
-
-	self.RecordInfo = recordinfo
-	self.Record = buf
-
 	if self.SPSInfo, err = ParseSPS(sps); err != nil {
 		return
 	}
+	buf := make([]byte, recordinfo.Len())
+	recordinfo.Marshal(buf, self.SPSInfo)
+	self.RecordInfo = recordinfo
+	self.Record = buf
 	return
 }
 
@@ -480,14 +512,12 @@ func (self *AVCDecoderConfRecord) Unmarshal(b []byte) (n int, err error) {
 		err = ErrDecconfInvalid
 		return
 	}
-
 	self.AVCProfileIndication = b[1]
 	self.ProfileCompatibility = b[2]
 	self.AVCLevelIndication = b[3]
 	self.LengthSizeMinusOne = b[4] & 0x03
 	spscount := int(b[5] & 0x1f)
 	n += 6
-
 	for i := 0; i < spscount; i++ {
 		if len(b) < n+2 {
 			err = ErrDecconfInvalid
@@ -550,55 +580,63 @@ func (self *AVCDecoderConfRecord) Unmarshal(b []byte) (n int, err error) {
 }
 
 func (self AVCDecoderConfRecord) Len() (n int) {
-	n = 7
+	n = 23
 	for _, sps := range self.SPS {
-		n += 2 + len(sps)
+		n += 5 + len(sps)
 	}
 	for _, pps := range self.PPS {
-		n += 2 + len(pps)
+		n += 5 + len(pps)
 	}
 	for _, vps := range self.VPS {
-		n += 2 + len(vps)
+		n += 5 + len(vps)
 	}
 	return
 }
 
-func (self AVCDecoderConfRecord) Marshal(b []byte) (n int) {
+func (self AVCDecoderConfRecord) Marshal(b []byte, si SPSInfo) (n int) {
 	b[0] = 1
 	b[1] = self.AVCProfileIndication
 	b[2] = self.ProfileCompatibility
 	b[3] = self.AVCLevelIndication
-	b[4] = self.LengthSizeMinusOne | 0xfc
+	b[21] = 3
-	b[5] = uint8(len(self.SPS)) | 0xe0
+	b[22] = 3
-	n += 6
+	n += 23
-
+	b[n] = (self.VPS[0][0] >> 1) & 0x3f
-	for _, sps := range self.SPS {
-		pio.PutU16BE(b[n:], uint16(len(sps)))
-		n += 2
-		copy(b[n:], sps)
-		n += len(sps)
-	}
-
-	b[n] = uint8(len(self.PPS))
 	n++
-
+	b[n] = byte(len(self.VPS) >> 8)
-	for _, pps := range self.PPS {
+	n++
-		pio.PutU16BE(b[n:], uint16(len(pps)))
+	b[n] = byte(len(self.VPS))
-		n += 2
-		copy(b[n:], pps)
-		n += len(pps)
-	}
-
-	b[n] = uint8(len(self.VPS))
 	n++
-
 	for _, vps := range self.VPS {
 		pio.PutU16BE(b[n:], uint16(len(vps)))
 		n += 2
 		copy(b[n:], vps)
 		n += len(vps)
 	}
-
+	b[n] = (self.SPS[0][0] >> 1) & 0x3f
+	n++
+	b[n] = byte(len(self.SPS) >> 8)
+	n++
+	b[n] = byte(len(self.SPS))
+	n++
+	for _, sps := range self.SPS {
+		pio.PutU16BE(b[n:], uint16(len(sps)))
+		n += 2
+		copy(b[n:], sps)
+		n += len(sps)
+	}
+	b[n] = (self.PPS[0][0] >> 1) & 0x3f
+	n++
+	b[n] = byte(len(self.PPS) >> 8)
+	n++
+	b[n] = byte(len(self.PPS))
+	n++
+	for _, pps := range self.PPS {
+		pio.PutU16BE(b[n:], uint16(len(pps)))
+		n += 2
+		copy(b[n:], pps)
+		n += len(pps)
+	}
 	return
 }
 
@@ -623,17 +661,13 @@ const (
 )
 
 func ParseSliceHeaderFromNALU(packet []byte) (sliceType SliceType, err error) {
-
 	if len(packet) <= 1 {
 		err = fmt.Errorf("h265parser: packet too short to parse slice header")
 		return
 	}
-
 	nal_unit_type := packet[0] & 0x1f
 	switch nal_unit_type {
 	case 1, 2, 5, 19:
-		// slice_layer_without_partitioning_rbsp
-		// slice_data_partition_a_layer_rbsp
 
 	default:
 		err = fmt.Errorf("h265parser: nal_unit_type=%d has no slice header", nal_unit_type)
@@ -641,13 +675,9 @@ func ParseSliceHeaderFromNALU(packet []byte) (sliceType SliceType, err error) {
 	}
 
 	r := &bits.GolombBitReader{R: bytes.NewReader(packet[1:])}
-
-	// first_mb_in_slice
 	if _, err = r.ReadExponentialGolombCode(); err != nil {
 		return
 	}
-
-	// slice_type
 	var u uint
 	if u, err = r.ReadExponentialGolombCode(); err != nil {
 		return
@@ -664,6 +694,5 @@ func ParseSliceHeaderFromNALU(packet []byte) (sliceType SliceType, err error) {
 		err = fmt.Errorf("h265parser: slice_type=%d invalid", u)
 		return
 	}
-
 	return
 }

format/mp4f/muxer.go

@@ -181,6 +181,7 @@ func (self *Stream) fillTrackAtom() (err error) {
 	} else if self.Type() == av.H265 {
 		codec := self.CodecData.(h265parser.CodecData)
 		width, height := codec.Width(), codec.Height()
+
 		self.sample.SampleDesc.HV1Desc = &mp4io.HV1Desc{
 			DataRefIdx:           1,
 			HorizontalResolution: 72,
@@ -192,6 +193,7 @@ func (self *Stream) fillTrackAtom() (err error) {
 			ColorTableId:         -1,
 			Conf:                 &mp4io.HV1Conf{Data: codec.AVCDecoderConfRecordBytes()},
 		}
+
 		self.trackAtom.Media.Handler = &mp4io.HandlerRefer{
 			SubType: [4]byte{'v', 'i', 'd', 'e'},
 			Name:    []byte{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 'G', 'G', 0, 0, 0},
@@ -200,6 +202,7 @@ func (self *Stream) fillTrackAtom() (err error) {
 			Flags: 0x000001,
 		}
 		self.codecString = fmt.Sprintf("hvc1.%02X%02X%02X", codec.RecordInfo.AVCProfileIndication, codec.RecordInfo.ProfileCompatibility, codec.RecordInfo.AVCLevelIndication)
+
 	} else if self.Type() == av.AAC {
 		codec := self.CodecData.(aacparser.CodecData)
 		self.sample.SampleDesc.MP4ADesc = &mp4io.MP4ADesc{
@@ -354,6 +357,7 @@ func (element *Stream) writePacketV3(pkt av.Packet, rawdur time.Duration, maxFra
 }
 
 func (element *Stream) writePacketV2(pkt av.Packet, rawdur time.Duration, maxFrames int) (bool, []byte, error) {
+	//pkt.Data = pkt.Data[4:]
 	trackID := pkt.Idx + 1
 	if element.sampleIndex == 0 {
 		element.moof.Header = &mp4fio.MovieFragHeader{Seqnum: uint32(element.muxer.fragmentIndex + 1)}

format/rtspv2/client.go

@@ -142,26 +142,33 @@ func Dial(options RTSPClientOptions) (*RTSPClient, error) {
 			return nil, err
 		}
 		if i2.AVType == VIDEO {
-			if i2.Type == av.H264 && len(i2.SpropParameterSets) > 1 {
+			if i2.Type == av.H264 {
+				if len(i2.SpropParameterSets) > 1 {
 					if codecData, err := h264parser.NewCodecDataFromSPSAndPPS(i2.SpropParameterSets[0], i2.SpropParameterSets[1]); err == nil {
 						client.sps = i2.SpropParameterSets[0]
 						client.pps = i2.SpropParameterSets[1]
 						client.CodecData = append(client.CodecData, codecData)
-					client.videoIDX = int8(len(client.CodecData) - 1)
-					client.videoCodec = av.H264
 					}
-			} else if i2.Type == av.H265 && len(i2.SpropVPS) > 1 && len(i2.SpropSPS) > 1 && len(i2.SpropPPS) > 1 {
+				} else {
+					client.CodecData = append(client.CodecData, h264parser.CodecData{})
+				}
+				client.videoCodec = av.H264
+			} else if i2.Type == av.H265 {
+				if len(i2.SpropVPS) > 1 && len(i2.SpropSPS) > 1 && len(i2.SpropPPS) > 1 {
 					if codecData, err := h265parser.NewCodecDataFromVPSAndSPSAndPPS(i2.SpropVPS, i2.SpropSPS, i2.SpropPPS); err == nil {
 						client.vps = i2.SpropVPS
 						client.sps = i2.SpropSPS
 						client.pps = i2.SpropPPS
 						client.CodecData = append(client.CodecData, codecData)
-					client.videoIDX = int8(len(client.CodecData) - 1)
-					client.videoCodec = av.H265
 					}
+				} else {
+					client.CodecData = append(client.CodecData, h265parser.CodecData{})
+				}
+				client.videoCodec = av.H265
 			} else {
 				client.Println("SDP Video Codec Type Not Supported", i2.Type)
 			}
+			client.videoIDX = int8(len(client.CodecData) - 1)
 			client.videoID = ch
 		}
 		if i2.AVType == AUDIO {
@@ -552,7 +559,7 @@ func (client *RTSPClient) RTPDemuxer(payloadRAW *[]byte) ([]*av.Packet, bool) {
 						Data:            append(binSize(len(nal)), nal...),
 						CompositionTime: time.Duration(1) * time.Millisecond,
 						Idx:             client.videoIDX,
-						IsKeyFrame:      naluType == h265parser.NAL_UNIT_CODED_SLICE_IDR_W_RADL,
+						IsKeyFrame:      false,
 						Duration:        time.Duration(float32(timestamp-client.PreVideoTS)/90) * time.Millisecond,
 						Time:            time.Duration(timestamp/90) * time.Millisecond,
 					})
@@ -564,22 +571,19 @@ func (client *RTSPClient) RTPDemuxer(payloadRAW *[]byte) ([]*av.Packet, bool) {
 					client.CodecUpdatePPS(nal)
 				case h265parser.NAL_UNIT_UNSPECIFIED_49:
 					se := nal[2] >> 6
-					naluType = nal[2] & 0x3f
+					naluType := nal[2] & 0x3f
 					if se == 2 {
 						client.BufferRtpPacket.Truncate(0)
 						client.BufferRtpPacket.Reset()
-						client.BufferRtpPacket.Write([]byte{0, 0, 0, 0, (nal[0] & 0x81) | (naluType << 1), nal[1]})
+						client.BufferRtpPacket.Write([]byte{(nal[0] & 0x81) | (naluType << 1), nal[1]})
 						r := make([]byte, 2)
 						r[1] = nal[1]
 						r[0] = (nal[0] & 0x81) | (naluType << 1)
 						client.BufferRtpPacket.Write(nal[3:])
 					} else if se == 1 {
 						client.BufferRtpPacket.Write(nal[3:])
-						binary.BigEndian.PutUint32(client.BufferRtpPacket.Bytes()[:4], uint32(client.BufferRtpPacket.Len())-4)
-						buf := make([]byte, client.BufferRtpPacket.Len())
-						copy(buf, client.BufferRtpPacket.Bytes())
 						retmap = append(retmap, &av.Packet{
-							Data:            append(binSize(len(nal)), nal...),
+							Data:            append(binSize(client.BufferRtpPacket.Len()), client.BufferRtpPacket.Bytes()...),
 							CompositionTime: time.Duration(1) * time.Millisecond,
 							Idx:             client.videoIDX,
 							IsKeyFrame:      naluType == h265parser.NAL_UNIT_CODED_SLICE_IDR_W_RADL,
@@ -590,8 +594,9 @@ func (client *RTSPClient) RTPDemuxer(payloadRAW *[]byte) ([]*av.Packet, bool) {
 						client.BufferRtpPacket.Write(nal[3:])
 					}
 				default:
-					//log.Println("???", naluType)
+					client.Println("Unsupported Nal", naluType)
 				}
+
 			} else if client.videoCodec == av.H264 {
 				naluType := nal[0] & 0x1f
 				switch {

format/webrtcv3/adapter.go

@@ -195,6 +195,9 @@ func (element *Muxer) WritePacket(pkt av.Packet) (err error) {
 	}
 	if tmp, ok := element.streams[pkt.Idx]; ok {
 		element.StreamACK.Reset(10 * time.Second)
+		if len(pkt.Data) < 5 {
+			return nil
+		}
 		switch tmp.codec.Type() {
 		case av.H264:
 			codec := tmp.codec.(h264parser.CodecData)

go.mod

@@ -5,4 +5,5 @@ go 1.14
 require (
 	github.com/pion/webrtc/v2 v2.2.23 // indirect
 	github.com/pion/webrtc/v3 v3.0.1 // indirect
+	github.com/q191201771/lal v0.19.1 // indirect
 )

go.sum

@@ -100,6 +100,10 @@ github.com/pion/webrtc/v3 v3.0.1/go.mod h1:ePzv8r2tzj95nJuPsns/7OiS5M8RiGSULdxzI
 github.com/pkg/errors v0.9.1 h1:FEBLx1zS214owpjy7qsBeixbURkuhQAwrK5UwLGTwt4=
 github.com/pkg/errors v0.9.1/go.mod h1:bwawxfHBFNV+L2hUp1rHADufV3IMtnDRdf1r5NINEl0=
 github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
+github.com/q191201771/lal v0.19.1 h1:Fxh33T+9y7Dp958/SFalzpWlt18LEYYT2z8UZdzhokQ=
+github.com/q191201771/lal v0.19.1/go.mod h1:HUvNYsaXA7ZwGStxtpnxpH0lN5qzd350glLmdigDg0A=
+github.com/q191201771/naza v0.17.0 h1:ZuETbHDX8srxxxL4AiDW+HTwEHnbfJvS40R8PUNStpQ=
+github.com/q191201771/naza v0.17.0/go.mod h1:5LeGupZZFtYP1g/S203n9vXoUNVdlRnPIfM6rExjqt0=
 github.com/sclevine/agouti v3.0.0+incompatible/go.mod h1:b4WX9W9L1sfQKXeJf1mUTLZKJ48R1S7H23Ji7oFO5Bw=
 github.com/stretchr/objx v0.1.0/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME=
 github.com/stretchr/testify v1.3.0/go.mod h1:M5WIy9Dh21IEIfnGCwXGc5bZfKNJtfHm1UVUgZn+9EI=

utils/bits/golomb_reader.go

@@ -33,6 +33,30 @@ func (self *GolombBitReader) ReadBits(n int) (res uint, err error) {
 	return
 }
 
+func (self *GolombBitReader) ReadBits32(n uint) (r uint32, err error) {
+	var t uint
+	for i := uint(0); i < n; i++ {
+		t, err = self.ReadBit()
+		if err != nil {
+			return
+		}
+		r = (r << 1) | uint32(t)
+	}
+	return
+}
+
+func (self *GolombBitReader) ReadBits64(n uint) (r uint64, err error) {
+	var t uint
+	for i := uint(0); i < n; i++ {
+		t, err = self.ReadBit()
+		if err != nil {
+			return
+		}
+		r = (r << 1) | uint64(t)
+	}
+	return
+}
+
 func (self *GolombBitReader) ReadExponentialGolombCode() (res uint, err error) {
 	i := 0
 	for {