We are developing an audio recording application for 10.7.5 and above.
We are using AVCaptureAudioDataOutput to capture audio and data is written using AVAssetWriterInput.
Properties set to the AVAssetWriterInput as below:
AVFormatIDKey : kAudioFormatMPEG4AAC
AVSampleRateKey : This values is taken from device using below code
AudioStreamBasicDescription streanDesc;
UInt32 propSize = sizeof(AudioStreamBasicDescription);
AudioDeviceGetProperty(mInputDevice, 0, YES, kAudioDevicePropertyStreamFormat, &propSize,&streanDesc);
AVNumberOfChannelsKey : This is taken from AudioStreamBasicDescription.
AVEncoderBitRateKey : This is hardcoded to 96000.
This works fine for most of the audio device except for USB mic of sample rate 32khz and iSight device with sample rate 48khz.
When we use these two devices as input audio device, while writing audio data we are getting the following error -
Error Domain=AVFoundationErrorDomain Code=-11821 "Cannot Decode" UserInfo=0x279470 {NSLocalizedFailureReason=The media data could not be decoded. It may be damaged., NSUnderlyingError=0x292370 "The operation couldn’t be completed. (OSStatus error 560226676.)", NSLocalizedDescription=Cannot Decode}".
However in case of USB mic if we hardcode the AVSampleRateKey to 44100 it works perfectly fine but does not work for iSight device.
What is correct value to be provided for AVSampleRateKey? Can any one help me to resolve this issue.
Related
I am generating an rtsp stream using gstreamer in an iOS app and trying to use ffmpeg in a Mac OS X audio driver that I wrote using XCode to strip the audio out of the stream and then pump the audio to Skype or Zoom or whatever. All the code is written in the 'C' old-fashioned programming language. I do get a FILE* that is not NULL back from a popen function call to execute ffmpeg on the input rtsp stream. But once I get that FILE* object and try to read binary data from it, it returns that zero bytes have been read. Here is the code:
FILE *readPipeFromFFMPEG = popen("Contents/MacOS/ffmpeg -i rtsp://192.168.0.30:8554/test -vn -acodec copy -flush_packets pipe:1", "r+");
int pipeFD = fileno(readPipeFromFFMPEG);
char *buffer = (char*)calloc(inIOBufferFrameSize * 8, 1);
numBytesRead = read(pipeFD, buffer, inIOBufferFrameSize * 8);
free(buffer);
pclose(readPipeFromFFMPEG);
but numBytesRead is always coming back as zero. Does anybody have any clue what I need to do to get this working properly? It seems like maybe a permissions issue where I do not have permission to read from the stream? Or maybe my ffmpeg parameters are incorrect? I am able to open up the stream in VLC and OBS Studio no problem and it displays the video frames and plays the audio. I am really stuck I need help! I must be missing something totally obvious because when I run OBS Studio or VLC it shows in the iPhone app the requests from the client because it prints out information that the audio and video packets are being requested but when the audio driver is running nothing is printed out in the iPhone app.
I'm trying to use various audio sources in DirectShow and I have these capture devices in my system which I think are quite common (provided by chipset drivers):
Realtek HD Audio Line input
Realtek HD Audio Stereo input
Realtek HD Audio Mic input
They look like capture sources, expose analog input and 24-bit pcm output, and can connect the output to other filters (renderer etc).
But the return code from IMediaFilter::Run of the capture filter is ERROR_BAD_COMMAND which does not say much. I tried it in my program and also in GraphStudioNext which did not reveal any extra information.
Is it possible to use these for capture and how?
Update
For instance, I tried this graph with mic input (actually connected and working). In this setup, the graph does not start (ERROR_BAD_COMMAND) but with the other source, it would start.
This is the same device but different drivers. The one that works is from the category "Audio capture sources" the one that does not "WDM Streaming Capture Devices".
The easiest way to check the device with GraphStudioNext is to build a recording graph with the PCM audio input device itself, AVI Mux filter and File Writer filter connected as this (with default media types):
You hit Run and the recording graph produces non-empty file via Filter Writer in the location prompted during graph building.
--
So now I realized your question is a bit different. You see filters corresponding to your audio input device both under
Audio Capture Sources -- CLSID_AudioInputDeviceCategory
WDM Streaming Capture Devices -- AM_KSCATEGORY_CAPTURE
And the question is that the first filter works and the other does not.
A similar filter from AM_KSCATEGORY_CAPTURE seems to be connecting into topology, but attempt to run triggers ERROR_BAD_COMMAND.
First of all, these are indeed different filters. Even though underlying hardware might be the same, the "frontend" filters are different. The wrapper that "works" is Audio Capture Filter backed by WDM device. In the other case it is Generic WDM Filter Proxy which behavior is, generally speaking, undefined. The filter is not documented and, I am guessing, it does not receive sufficient initialization or does not implement required behavior otherwise, so this proxy is not and is not supposed to be interchangeable with Audio Capture Filter proxy.
I am processing a video file.
I use ffmpeg to read each packet.
If it is an audio packet, I write the packet into the output video file using av_interleaved_write_frame.
If it is a video packet, I decode the packet, get the data of the video frame, process the image, and compress back to a packet. Then I write the processed video frame packet into the output video file using av_interleaved_write_frame.
Through debugging, it read audio packets and video packets correctly.
However, when it goes to "av_write_trailer", it exits. But the output video file exists.
The error information is:
*** glibc detected *** /OpenCV_videoFlatten_20130507/Debug/OpenCV_videoFlatten_20130507: corrupted double-linked list: 0x000000000348dfa0 ***
Using Movie Player (in Ubuntu), the output video file can plays the audio correctly, but without video signals.
Using VLC player, it can show the first video frame (keep the same video picture), and play the audio correctly.
I tried to debug into "av_write_trailer", but since it is in the ffmpeg library, I could not get a detailed information what is wrong.
Another piece of information: the previous version of the project is only to process the video frame, without adding audio stream; and it works well.
Any hint or clue?
I found the solution. I did not use rescale to set the pts based on stream's time_base. Actually the related code is in the example muxing.c.
i'm currently reading the documentation of MSDN to render a stream to an audio renderer..
or in other word, to play my captured data from microphone.
http://msdn.microsoft.com/en-us/library/dd316756%28v=vs.85%29.aspx
this example provides example.
My problem now is that i couldn't really understand the project flow.
I currently have a different class storing the below parameters which i obtained from the capture process.
these parameters will be continuously re-written as the program captures streaming audio data from the microphone.
BYTE data;
UINT32 bufferframecount;
DWORD flag;
WAVEFORMATEX *pwfx;
My question is,
How does really the loadData() function works.
is it suppose to grab the parameter i'm writing from capture process?
how does the program sends the data to audio renderer, and play it in my speaker.
The loadData() function fills in audio pointed to by pData. The example abstracts the audio source, so this could be anything from a .wav file to the microphone audio that you already captured.
So, if you are trying to build from that example, I would implement the MyAudioSource class, and have it just read PCM or float samples from a file, whenever loadData() is called. Then, if you run that program, it should play the audio from the file out the speaker.
Kodak Playsport camera HDMI output to Canopus ADVC HD50. (works okay for iMovie)
Need source code for a simple program similar to Apples' "MyRecorder" tutorial that demonstrates video capture, and then a report of what resolution it is.
Currently my program reports: "The operation couldn’t be completed. (OSStatus error -8961.)". Make sure that the formats of all video outputs are properly configured."
The answer for me is to purchase Calibrated {Q}'s XD Decode codec.