Fix pix_fmt for publishing to RTMP servers

README.md

@@ -779,7 +779,7 @@ POST http://localhost:1984/api/streams?dst=camera1&src=ffmpeg:http://example.com
 You can publish any stream to streaming services (YouTube, Telegram, etc.) via RTMP/RTMPS. Important:
 
 - Supported codecs: H264 for video and AAC for audio
-- Pixel format should be `yuv420p`, for cameras with `yuvj420p` format you SHOULD use [transcoding](#source-ffmpeg)
+- AAC audio is required for YouTube, videos without audio will not work
 - You don't need to enable [RTMP module](#module-rtmp) listening for this task
 
 You can use API:
@@ -792,16 +792,19 @@ Or config file:
 
 ```yaml
 publish:
-  # publish stream "tplink_tapo" to Telegram
-  tplink_tapo: rtmps://xxx-x.rtmp.t.me/s/xxxxxxxxxx:xxxxxxxxxxxxxxxxxxxxxx
-  # publish stream "other_camera" to Telegram and YouTube
-  other_camera:
+  # publish stream "video_audio_transcode" to Telegram
+  video_audio_transcode:
     - rtmps://xxx-x.rtmp.t.me/s/xxxxxxxxxx:xxxxxxxxxxxxxxxxxxxxxx
-    - rtmps://xxx.rtmp.youtube.com/live2/xxxx-xxxx-xxxx-xxxx-xxxx
+  # publish stream "audio_transcode" to Telegram and YouTube
+  audio_transcode:
+    - rtmps://xxx-x.rtmp.t.me/s/xxxxxxxxxx:xxxxxxxxxxxxxxxxxxxxxx
+    - rtmp://xxx.rtmp.youtube.com/live2/xxxx-xxxx-xxxx-xxxx-xxxx
 
 streams:
-  # for TP-Link cameras it's important to use transcoding because of wrong pixel format
-  tplink_tapo: ffmpeg:rtsp://user:pass@192.168.1.123/stream1#video=h264#hardware#audio=aac
+  video_audio_transcode:
+    - ffmpeg:rtsp://user:pass@192.168.1.123/stream1#video=h264#hardware#audio=aac
+  audio_transcode:
+    - ffmpeg:rtsp://user:pass@192.168.1.123/stream1#video=copy#audio=aac
 ```
 
 - **Telegram Desktop App** > Any public or private channel or group (where you admin) > Live stream > Start with... > Start streaming.

pkg/bits/reader.go

@@ -131,3 +131,7 @@ func (r *Reader) ReadSEGolomb() int32 {
 func (r *Reader) Left() []byte {
 	return r.buf[r.pos:]
 }
+
+func (r *Reader) Pos() (int, byte) {
+	return r.pos - 1, r.bits
+}

pkg/flv/muxer.go

@@ -54,6 +54,8 @@ func (m *Muxer) GetInit() []byte {
 			sps, pps := h264.GetParameterSet(codec.FmtpLine)
 			if len(sps) == 0 {
 				sps = []byte{0x67, 0x42, 0x00, 0x0a, 0xf8, 0x41, 0xa2}
+			} else {
+				h264.FixPixFmt(sps)
 			}
 			if len(pps) == 0 {
 				pps = []byte{0x68, 0xce, 0x38, 0x80}

pkg/h264/sps.go

@@ -1,6 +1,10 @@
 package h264
 
-import "github.com/AlexxIT/go2rtc/pkg/bits"
+import (
+	"fmt"
+
+	"github.com/AlexxIT/go2rtc/pkg/bits"
+)
 
 // http://www.itu.int/rec/T-REC-H.264
 // https://webrtc.googlesource.com/src/+/refs/heads/main/common_video/h264/sps_parser.cc
@@ -229,3 +233,132 @@ func (s *SPS) scaling_list(r *bits.Reader, sizeOfScalingList int) {
 		}
 	}
 }
+
+func (s *SPS) Profile() string {
+	switch s.profile_idc {
+	case 0x42:
+		return "Baseline"
+	case 0x4D:
+		return "Main"
+	case 0x58:
+		return "Extended"
+	case 0x64:
+		return "High"
+	}
+	return fmt.Sprintf("0x%02X", s.profile_idc)
+}
+
+func (s *SPS) PixFmt() string {
+	if s.bit_depth_luma_minus8 == 0 {
+		switch s.chroma_format_idc {
+		case 1:
+			if s.video_full_range_flag == 1 {
+				return "yuvj420p"
+			}
+			return "yuv420p"
+		case 2:
+			return "yuv422p"
+		case 3:
+			return "yuv444p"
+		}
+	}
+	return ""
+}
+
+func (s *SPS) String() string {
+	return fmt.Sprintf(
+		"%s %d.%d, %s, %dx%d",
+		s.Profile(), s.level_idc/10, s.level_idc%10, s.PixFmt(), s.Width(), s.Height(),
+	)
+}
+
+// FixPixFmt - change yuvj420p to yuv420p in SPS
+// same as "-c:v copy -bsf:v h264_metadata=video_full_range_flag=0"
+func FixPixFmt(sps []byte) {
+	r := bits.NewReader(sps)
+
+	_ = r.ReadByte()
+
+	profile := r.ReadByte()
+	_ = r.ReadByte()
+	_ = r.ReadByte()
+	_ = r.ReadUEGolomb()
+
+	switch profile {
+	case 100, 110, 122, 244, 44, 83, 86, 118, 128, 138, 139, 134, 135:
+		n := byte(8)
+
+		if r.ReadUEGolomb() == 3 {
+			_ = r.ReadBit()
+			n = 12
+		}
+
+		_ = r.ReadUEGolomb()
+		_ = r.ReadUEGolomb()
+		_ = r.ReadBit()
+
+		if r.ReadBit() != 0 {
+			for i := byte(0); i < n; i++ {
+				if r.ReadBit() != 0 {
+					return // skip
+				}
+			}
+		}
+	}
+
+	_ = r.ReadUEGolomb()
+
+	switch r.ReadUEGolomb() {
+	case 0:
+		_ = r.ReadUEGolomb()
+	case 1:
+		_ = r.ReadBit()
+		_ = r.ReadSEGolomb()
+		_ = r.ReadSEGolomb()
+
+		n := r.ReadUEGolomb()
+		for i := uint32(0); i < n; i++ {
+			_ = r.ReadSEGolomb()
+		}
+	}
+
+	_ = r.ReadUEGolomb()
+	_ = r.ReadBit()
+
+	_ = r.ReadUEGolomb()
+	_ = r.ReadUEGolomb()
+
+	if r.ReadBit() == 0 {
+		_ = r.ReadBit()
+	}
+
+	_ = r.ReadBit()
+
+	if r.ReadBit() != 0 {
+		_ = r.ReadUEGolomb()
+		_ = r.ReadUEGolomb()
+		_ = r.ReadUEGolomb()
+		_ = r.ReadUEGolomb()
+	}
+
+	if r.ReadBit() != 0 {
+		if r.ReadBit() != 0 {
+			if r.ReadByte() == 255 {
+				_ = r.ReadUint16()
+				_ = r.ReadUint16()
+			}
+		}
+
+		if r.ReadBit() != 0 {
+			_ = r.ReadBit()
+		}
+
+		if r.ReadBit() != 0 {
+			_ = r.ReadBits8(3)
+			if r.ReadBit() == 1 {
+				pos, bit := r.Pos()
+				sps[pos] &= ^byte(1 << bit)
+			}
+		}
+	}
+}