Flv structure description

1. flv

• 1. flv
  • 1.1. brief introduction
  • 1.2. structure
    • 1.2.1. The overall structure
    • 1.2.2. Header structure
    • 1.2.3. tag structure
    • 1.2.4. Audio data structure
    • 1.2.5. Video data structure
  • 1.3. Relevant concepts
  • 1.4. Reference material

1.1. brief introduction

• One flv file , Every type of tag All belong to one stream ( Or audio , Or video , Or information )
• One flv The file has at most one audio stream , A video stream , There is no case of multiple independent audio and video streams in one file .（mp4 It supports multiple audio and video ）
• flv The file format uses the big endian .

1.2. structure

• notes ： The following data type in
  • UI Represents an unsigned integer , The numbers that follow indicate the number of digits in length . such as UI8, Represents an unsigned integer , The length is one byte .
  • UI24 It's three bytes .
  • UB Represents a bit field , such as UB5 Represents a byte of 5 position .

1.2.1. The overall structure

• The overall structure is as follows , from Header And Tag form
• Every Tag There is a piece under the 4bytes Space , Used to record this tag The length of ( Not included tagSize The amount of 4 Bytes ), Post can be used for reverse read processing (TagHeader Li Hui records Tag The length of , It can be used for forward reading data processing ).
• body contain Tag0Size+Tag1(Tag1Size)+…+TagN(TagNSize)
• tagSize = 11*8(tagHeaderSize)+tagDataSize(tagHeader Are there in )

Header
Tag0Size( No, Tag0 So here is 0)
Tag1
Tag1Size
Tag2
Tag2Size
…
…
TagN(lastTag)
TagNSize(lastTagSize)

1.2.2. Header structure

Field	type	Comment
Signature	UI	’F’(0X46)
Signature	UI8	‘L’(0X4C)
Signature	UI8	‘V’(0x56)
edition	UI8	FLV Version of .0x01 Express FLV The version is 1
Keep field	UB5	The top five must be 0
Is there an audio stream	UB1	Whether the audio stream has a flag
Keep field	UB1	Must be 0
Is there a video stream	UB1	Whether there is a flag in the video stream
File header size	UI32	FLV The size of the head (FLV Version is 1 Fixed as 9). The size of the file header here ( Including these four bytes )

1.2.3. tag structure

Field	type	Comment
TAG type	UI8	8: audio,9: video,18: script data( Description information ).
data size	UI24	The size of the data area , barring tag head . The total size of Baotou is 11 Bytes .
Time stamp	UI24	Current frame timestamp , In milliseconds . be relative to FLV The first of the files TAG timestamp . first tag The timestamp of is always 0, the truth is that dts
Timestamp extension field	UI8	If the timestamp is greater than 0xFFFFFF, This byte will be used . This byte is the high of the timestamp 8 position , The upper three bytes are low 24 position .
flow ID	U24	Always 0
Data area	UI8[n]	Distinguish by type , Including video , Audio , Description information

1.2.4. Audio data structure

Field	type	Comment
Audio format	UB4	0： Linear PCM, platform endian;1：ADPCM;2：MP3;3：Linear PCM, little endian;4：Nellymoser 16-kHz mono;5：Nellymoser 8-kHz mono;6：Nellymoser;7：G.711 A-law logarithmic PCM;8：G.711 mu-law logarithmic PCM;10：AAC;11：Speex;14：MP3 8-Khz;15：Device-specific sound;9,12,13：reserved;
Sampling rate	UB2	0：5.5-kHz;1：11-kHz;2：22-kHz;3：44-kHz(For AAC: always 3)
Sample size	UB1	0：snd8Bit;1：snd16Bit. Only uncompressed formats are valid , The compression format is in accordance with 16bit Sampling processing
channel	UB1	0： Mono ;1： stereo , Two channel .Nellymoser Must be 0,AAC Must be 1
Voice data	UI8[N]	If it is PCM Linear data , Every time you store 16bit Small end storage , A signed . If the audio format is AAC, The stored data is AAC AUDIO DATA( See the following for the specific format ), Otherwise, it is a linear array .

AACAUDIODATA

Field	Type	Comment
AACPacketType	U8	0: AAC sequence header;1: AAC raw
Data	UI8[n]	AudioSpecificConfig(ISO 14496-3) or Raw AAC frame data

1.2.5. Video data structure

Field	type	Comment
The frame type	UB4	1: keyframe (for AVC, a seekable frame)——h264 Of IDR, Keyframes , Reentrant frame ; 2: inter frame (for AVC, a non- seekable frame),h264 Normal frame ; 3: disposable inter frame (H.263 only);4: generated keyframe (reserved for server use only); 5: video info/command frame( For example, whether the logo is Seek etc. )
code ID	UB4	1: JPEG (currently unused); 2: Sorenson H.263;3: Screen video;4: On2 VP6;5: On2 VP6 with alpha channel; 6: Screen video version 2;7: AVC
Video data	UI[N]	Different codes , The format is different ,H263VIDEOPACKET/SCREENVIDEOPACKET/VP6FLVVIDEOPACKET/VP6FLVALPHAVIDEOPACKET/SCREENV2VIDEOPACKET/AVCVIDEOPACKET One of the structures of , Refer to the standard for details

AVCVIDEOPACKET

Field	type	Comment
AVC packet type	UI8	0：AVC Sequence header 1：AVC NALU unit 2：AVC End of sequence ( Low level avc Unwanted )
CTS	SI24	If AVC packet The type is 1, Then for cts(Composition time offset) The offset , Otherwise 0, adopt CTS and tag Medium DTS You can get PTS
data	UI8[n]	If AVC packet The type is 0, Is the decoder configuration ,sps,pps. If it is 1, It is nalu unit （ It can be more than one ）

1.3. Relevant concepts

• pts（presentation time stamps）,dts(decoder timestamps),cts(CompositionTime)
  • pts： Display time , That is, the time when the receiver displays this frame on the display . Unit is 1/90000 second .
  • dts： Decoding time , That is to say rtp The timestamp transmitted in the packet , Indicates the decoding order . The unit is 1/90000 second .
  • cts The offset ：cts = (pts - dts) / 90 .cts In milliseconds .
  • flv In file Timestamp and TimestampExtended What you spell is dts. Unit is ms.
  • CompositionTime Express PTS be relative to DTS Offset value , Display time (pts) = Decoding time （tag Of the 5~8 byte ） + CompositionTime, The unit is also ms

1.4. Reference material

1. FLV Detailed explanation of official specification of document format
2. Official documents
3. flv Medium time stamp