I'm working with libffmpeg in an iOS app. My goal is to connect to an RTSP source and write the media out to a file that can later be used with the iOS media player. Ideally I'd like to do this without transcoding the incoming data. I also want to be able to later re-encode the media with AVAssetExportSession if the user chooses to do so.
Because I want to create a file that is compatible with iOS, I'm limited (I believe) to mpeg, mp4 or quicktime (mov) formats.
Whenever I try to use one of these formats, I see the following warnings during my call to avformat_write_header:
[mov # 0x16401c00] Codec for stream 0 does not use global headers but container format requires global headers
[mov # 0x16401c00] Codec for stream 1 does not use global headers but container format requires global headers
My understanding is that the header wants to know the ultimate file size, which I do not know (the RTSP server is live streaming a camera, and the user stops the recording whenever they want). I guess that makes sense, but I know that others have successfully done this using the ffmpeg command line, so I'm confused as to what else I need to do here.
If I ignore the warning, I can still proceed with writing the file. If I choose mpeg or mp4 formats, my app crashes when I call av_write_trailer. If I use mov, I can successfully close the file, and the file does play back, but usually fails when I try to hand it to the AVAssetExportSession.
I would appreciate any insight into this. Thanks.
Frank
I found what appears to be a solution -- at least, it eliminates the warning. I had to set the CODEC_FLAG_GLOBAL_HEADER on both the audio and video codecs, before calling avcodec_open2.
Related
i have video uncompressed .avi video files. however they come with a known 2048kb header at the beginning of the file, such that no video program (Vlc player) recognises the file as video files. can i force VLC to ignore the first 2048kb? how?
what would be the easiest way to remove this header?
Unfortunately there is no known to me way of getting vlc to use a larger "probesize" or somehow just skip to byte X.
Also unfortunately there is no known to me way to add some "camouflage" or even add an empty chunk to the front of an avi container. Next approach could be to genereate a reference movie but also no automatic usage tools available at the best of my knowledge. And as there is also no way to just "set or change" the start offset of a file, we have to copy it.
My recommendation for copying parts of files is this sourceforge project: skf
https://sourceforge.net/projects/swissfileknife/?source=typ_redirect
some examples
I've been asked by my client whether it is possible to download a video and stream it once a bit has downloaded, just like pocketcasts does. His reasoning is this will allow him to store his video files on a site such as godaddy and bypass the need to stream the file to the phone which normally requires a dedicated server.
Is this even possible? if so do you know anywhere I can look to find out how pocketcasts does it? At the moment my app just streams an mp4.
Thanks for looking,
Matt
Since you're targetting iOS, HLS (HTTP Live Streaming) is your friend: https://developer.apple.com/streaming/
Please see my answer here for how you can use it: Simultaneously downloading and playing a song that is pieced together from multiple URLs
It's very easy to run a long movie through the mediafilesegmenter tool from Apple (or FFMPEG) which spits out a number of small .ts files (MPEG 2 Transport Stream). Then you create a manifest (a .m3u8 file) which describes how these files fit together (which mediafilesegment will create for you too!). Then you just put the manifest file and the .ts files on a hosting provider (like GoDaddy) and you're all set.
For example, given a file called test.mp4, first turn it into a .ts file with ffmpeg:
ffmpeg -i test.mp4 -acodec copy -vcodec copy -bsf h264_mp4toannexb test.ts
Then turn it into a series of HLS segments with mediafilesegmenter (the same can be done using the ffmpeg segment muxer, but mediafilesegmenter seems to be more robust):
mediafilesegmenter -t 3 test.ts
The result is a bunch of 3 second clips (that's what -t 3 means) and an manifest file called prog_index.m3u8. The contents of that look like:
#EXTM3U
#EXT-X-TARGETDURATION:3
#EXT-X-VERSION:3
#EXT-X-MEDIA-SEQUENCE:0
#EXT-X-PLAYLIST-TYPE:VOD
#EXTINF:2.99520,
fileSequence0.ts
#EXTINF:2.99520,
fileSequence1.ts
#EXTINF:2.99520,
fileSequence2.ts
#EXTINF:2.99520,
fileSequence3.ts
...
#EXTINF:0.37440,
fileSequence75.ts
#EXT-X-ENDLIST
Simply putting all of the .ts files and the .m3u8 file on a web server and pointing your AVPlayer or MPMoviePlayerController in iOS at the URL for the .m3u8 will get you an excellent streaming performance.
I'm developing some Video Editing Apps on Android.
the objective of the app is "Editing Videos on Android".
and...
I'm just completed making video file using some images.
but.. I can't attach audio into the video.
my method is same as follows.
1.VideoStream, audio stream creation using AVFormatContext
2.Movie encoding in video stream was successful
3.Encode codec open in audio stream was successful
4.Set sample format to AV_SAMPLE_FMT_FLTP
5.Sample rate and channel was set same as source audio
6.Choose appropriate Decoder and read packet
7.Convert packets using swr_converter, setting same as sample format
8.Encode converted data
9.memory deallocation
10.END!
Problem is here:
Video of finally created video file was normally played. but the Audio wasn't.
It heared like weird. It have many noises and plays slowly.
I've googled with many keywords but they only say about "FFmpeg command line usage".
I wanna make with FFMpeg API. not a Command line tool.
Please help.
Your question is vague without some kind of code to go along with it, as trust me there are a lot of things that can go wrong when using ffmpeg's libraries directly (and on Windows there is no debuging). Unfortunately ffmpeg's libraries are not well documented so it is generally best to read the source code for ffmpeg in order to use its libraries. Find the equivalent command line options to perform what you want and track that through ffmpeg's source to see the library calls.
i am sending live rtmp stream to wowza server with a live application config but everytime i connect to the stream for watching the live stream, its start from the beginning of the stream. I can see wowza is creating bigger and bigger file in /content directory and this file will everytime be played from the beginning.
How can i say wowza to send it live, like send the last 10 seconds of the file?
Best regards,
Chris
You'll probably have more luck asking this question on the Wowza forums, where their support team regularly addresses these questions. You'll need to provide more information:
What is your input (camera, flash media encoder, file)?
Can you stream VOD?
A large file building in the content directory sounds like you may be recording the stream. Do you have any add-ons such as the live stream recorder installed?
What url are you using to connect?
Well that answer is simple. You set the wrong stream type in the config.
I guess you set rtp-live-record instead of rtp-live that should fix it.
Kind Regards, Sui
Follow these instructions to create a live stream and broadcasting it
Go to the Wowza Media Server Directory. Probably it is /usr/local/WowzaMediaServer
cd applications;mkdir live
Here live is your application name
cd ../conf;mkdir live;cp Application.xml live/
Now edit the Application.xml file
cd live;vim Application.xml
Change the stream type default to live
live
set HTTPStreamers
cupertinostreaming,smoothstreaming,sanjosestreaming
set LiveStreamPacketizers to
cupertinostreamingpacketizer, smoothstreamingpacketizer, sanjosestreamingpacketizer
set Playmethod
none
Now go to your browser and open the player
file:///usr/local/WowzaMediaServer/examples/LiveVideoStreaming/client/live.html
Server : rtmp://;1935/live
Stream :
Note * : For live streaming you have to use an encoder ( Adobe Flash media Live Encoder , etc )
I have a PHP MP3 file which changes every time it is loaded, and plays a different advert, what I would like is away or is there away to play that before the SHOUTcast stream as the SHOUTcast intro file.
I personally dont really want to put it into flash or create custom playlist files but more would like SHOUTcast to fetch an intro file from a URL address.
Unfortunately, the SHOUTcast configuration doesn't allow this directly.
However, you can easily accomplish what you need by having a script pick a new intro file every few seconds or so. Then you can either write that MP3 to the file configured in SHOUTcast, or create a symlink and update it.