I am trying to create snapshots from a video stream using the "scene" video filter. I'm on Windows for now, but this will run on Linux I don't want the video output window to display. I can get the scenes to generate if I don't use the --vout=dummy option. When I include that option, it does not generate the scenes.
This example on the Wiki indicates that it's possible. What am I doing wrong?
Here is the line of code from the LibVLCSharp code:
LibVLC libVLC = new LibVLC("--no-audio", "--no-spu", "--vout=dummy", "--video-filter=scene", "--scene-format=jpeg", "--scene-prefix=snap", "--scene-path=C:\\temp\\", "--scene-ratio=100", $"--rtsp-user={rtspUser}", $"--rtsp-pwd={rtspPassword}");
For VLC 3, you will need to disable hardware acceleration which seems incompatible with the dummy vout.
In my tests, it was needed to do that on the media rather than globally:
media.AddOption(":avcodec-hw=none");
I still have mainy "Too high level or recursion" errors, and for that, I guess you'd better open an issue on videolan's trac.
Related
I would like to automute certain audio clients when I want, without human intervention (e.g. pulling up pavucontrol) when using Fedora/pulsewire. This worked before using pacmd, but that doesn't work under pipewire; and the replacement command, pw-cli, doesn't support set-sink-input-mute or set-source-output-mute.
pw-cli doesn't seem to support muting that I could see. pw-mon shows a relevant change when using pavucontrol (or easyeffects) to mute a stream, but that didn't help me figure out how to do it myself.
This is a bit late but i recently needed it too. Instead of pacmd you can use pactl set-sink-input-volume <sink-id> <volume>
I'm currently making an electron app that needs to play some 40Mbyte audio file from the file system, maybe it's wrong to do this but I found that the only way to play from anywhere in the file system is to convert the file to a dataurl in the background script and then transfer it using icp, after that I simply do
this.sound = new Audio(dataurl);
this.sound.preload = "metadata"
this.sound.play()
(part of a VueJS component hence the this)
I did a profling inside electron and this is what came out:
Note that actually transferring the 40Mbytes audio file doesn't take that long (around 80ms) what is extremely annoying is the "Second Task" which is probably buffering (I have no idea) which last around 950ms, this is way too long and ideally would need it under <220ms
I've already tried changing the preload option to all available options and while I'm using the native html5 audio right now I've also tried howlerjs with similar results (seemed a bit faster tho).
I would guess that loading the file directly might be faster but even after disabling security measures put by electron to block the file:/// it isn't recognized as a valid URI by XHR
Is there a faster way to load the dataurl since all the data is there it just needs to be converted to a buffer or something like that ?
Note: I can not "pre-buffer" every file in advance since there is about 200 of them it just wouldn't make sense in my opinion.
Update:
I found this post Electron - throws Not allowed to load local resource when using showOpenDialog
don't know how I missed it, so I followed step 1 and I now can load files inside electron with the custom protocol, however, nor Audio nor howlerjs is faster, it's actually slower at around 6secs from click to first sound, is it that it needs to buffer the whole file before playing ?
Update 2:
It appears that the 6sec loading time is only effective on the first instance of audio that is created. I do not know why tho. After that the use of two instances (one playing and one pre-buffering) work just fine, however even loading a file that isn't loaded is instantaneous. Seems weird that it only is the firs one.
I am trying to get started using Haskell's Euterpea library. My first goal was to get it to play a given sound file (e.g. mp3 or wav), but first I ran into an issue following instructions to get it to just play a simple note sound in ghci.
Following the "Setting up MIDI" instructions at Euterpea's web page, I ran
import Euterpea
play $ c 4 qn
in ghci. The 'play' command resulted in the following error message:
Prelude Euterpea> play $ c 4 qn
*** Exception: No MIDI output device found
CallStack (from HasCallStack):
error, called at ./Euterpea/IO/MIDI/MidiIO.lhs:122:18 in Euterpea-2.0.2-Iz37iWlkpjn2emP4FnvOI1:Euterpea.IO.MIDI.MidiIO
I thought I needed to specify midi output to my machine (macOS Sierra) and found an application called 'Audio MIDI Setup', but it showed that a midi output (my internal speakers) was already specified.
Anyone know what this issue is or how to fix it?
Perhaps you solved this, but for posterity some ideas:
Sounds like you didn't install and run a MIDI synth (e.g. SimpleSynth) first. AFAICT, Audio MIDI Setup doesn't actually include a software synthesizer, it's more for advanced / hardware MIDI setup.
This should create the MIDI output devices that Euterpea couldn't find there. You may also need to play around with channels (e.g. use playDev n instead of play and work out a value for n from your device list... or even just try 1 through 8)
Either way there's some good Mac-focussed help on Donya's working with MIDI on Mac OS X page. Hope that helps.
I am currently creating a music player in python 3.3 and I have a way of opening the mp3/wav files, namely through using through 'os.startfile()', but, this way of running the files means that if I run more than one, the second cancels the first, and the third cancels the second, and so on and so forth, so I only end up running the last file. So, basically, I would like a way of reading the mp3 file length so that I can use 'time.sleep(SongLength)' between the start of each file.
Thanks in advance.
EDIT:
I forgot to mention, but I would prefer to do this using only pre-installed libraries, as i am hoping to publish this online as a part of a (much) larger program
i've managed to do this Using an external module, as after ages of trying to do it without any, i gave up and used tinytag, as it is easy to install and use.
Nothing you can do without external libraries, as far as I know. Try using pymad.
Use it like this:
import mad
SongFile = mad.MadFile("something.mp3")
SongLength = SongFile.total_time()
I use the audio class to read MP3 file thanks to a little trick : replacing the ffmpegsumo.so of Node-Webkit with the chromium one. This enable the reading of MP3 on Windows but doesn't works on Mac OS. Does anyone know why ?
Here's the code :
player = new Audio()
player.src = '/path/to/the/audio.mp3';
player.play();
This seems to be dependant upon the dll/so being a 32 bit version. I am guessing that is why copying the file from Chrome doesn't work correctly for most people ( my 3 year old phone is the only 32-bit device I have left ).
I keep seeing this link --
https://github.com/rogerwang/node-webkit/wiki/Support-mp3-and-h264-in-video-and-audio-tag
.. but it is a blank page. I am guessing it was deleted since the info was likely not current or correct.
This issue thread has links to some rebuilt ffmpegsumo libraries for both Mac and Windows --
https://github.com/rogerwang/node-webkit/issues/1423
The alternative appears to be rebuilding ffmpegsumo, this thread has some config for doing that -- https://github.com/rogerwang/node-webkit/issues/1208
I am still confused about the licensing on it after you build the library, so that is probably worth some research. Everything about mpeg4-part10 is copyrighted and heavily patent encumbered. I think we all need to get smart enough to stop using mp4/h.264. Before I got this working correctly on node-webkit, it was easier to use ffmpeg to transcode the video to an ogv container using Theora and Vorbis codecs. At this point it seems like iOS is keeping h.264 alive, when it should probably die the horrible death it has earned.