I've written an MP4 parser that can read atoms in an MP4 just fine, and stitch them back together - the result is a technically valid MP4 file that Quicktime can open and such, but it can't play any audio as I believe the timing/sampling information is all off. I should probably mention I'm only interested in audio.
What I'm doing is trying to take the moov atoms/etc from an existing MP4, and then take only a subset of the mdat atom in the file to create a new, smaller MP4. In doing so I've altered the duration in the mvhd atom, as well as the duration in the mdia header. There are no tkhd atoms in this file that have edits, so I believe I don't need to alter the durations there - what am I missing?
In creating the new MP4 I'm properly sectioning the mdat block with a wide box, and keeping the 'mdat' header/size in their right places - I make sure to update the size with the new content.
Now it's entirely 110% possible I'm missing something crucial about the format, but if this is possible I'd love to get the final piece. Anybody got any input/ideas?
Code can be found at the following link:
https://gist.github.com/ryanmcgrath/958c602cff133bd7fa0b
I'm going to take a stab in the dark here and say that you're not updating your stbl offsets properly. At least I didn't (at first glance) see your python doing that anywhere.
STSC
Lets start with the location of data. Packets are written into the file in terms of chunks, and the header tells the decoder where each "block" of these chunks exists. The stsc table says how many items per chunk exist. The first chunk says where that new chunk starts. It's a little confusing, but look at my example. This is saying that you have 100 samples per chunkk, up to the 8th chunk. At the 8th chunk there are 98 samples.
STCO
That said, you also have to track where the offsets of these chunks are. That's the job of the stco table. So, where in the file is chunk offset 1, or chunk offset 2, etc.
If you modify any data in mdat you have to maintain these tables. You can't just chop mdat data out, and expect the decoder to know what to do.
As if this wasn't enough, now you have to also maintain the sample time table (stts) the sample size table (stsz) and if this was video, the sync sample table (stss).
STTS
stts says how long a sample should play for in units of the timescale. If you're doing audio the timescale is probably 44100 or 48000 (kHz).
If you've lopped off some data, now everything could potentially be out of sync. If all the values here have the exact same duration though you'd be OK.
STSZ
stsz says what size each sample is in bytes. This is important for the decoder to be able to start at a chunk, and then go through each sample by its size.
Again, if all the sample sizes are exactly the same you'd be OK. Audio tends to be pretty much the same, but video stuff varies a lot (with keyframes and whatnot)
STSS
And last but not least we have the stss table which says which frame's are keyframes. I only have experience with AAC, but every audio frame is considered a keyframe. In that case you can have one entry that describes all the packets.
In relation to your original question, the time display isn't always honored the same way in each player. The most accurate way is to sum up the durations of all the frames in the header and use that as the total time. Other players use the metadata in the track headers. I've found it best to just keep all the values the same and then players are happy.
If you're doing all that and I missed it in the script then can you post a sample mp4 and a standalone app and I can try to help you out.
Related
I am writing my own Opus Ogg writer following these specifications: RFC7845 and RFC3533.
Currently, I am facing an issue that I believe is related to how I am setting the lacing values (segment table).
My current setup is to basically read (using an existing Ogg reader) an Ogg file with a single Opus track and put that Opus track in another Ogg file that I create using my own Ogg writer.
So I have a function that takes the Opus content of each page from the original Ogg file and put it in pages in my new Ogg file.
I am being able to create the file successfully, but when I try playing it on VLC, it shows the correct timestamp but it does not play any sound.
I noticed that the issue is being caused by the way my segment table (or lacing values) is set.
I am currently creating it by filling each segment with as much data as possible (i.e 255 bytes), and letting only the last segment have a size < 255. This seems to be the way that other implementations are doing it (see Rust implementation, C implementation).
However, when I inspect the lacing values for a page containing that Opus content in the original Ogg file, it is not filled with 255s. It's another combination of segment sizes that still sums up to the same page size, but that uses more segments (since it's not taking up the max segment size). When I try using the exact segments combination in the original file, the file plays on VLC successfully.
So that makes me conclude that the approach I am taking with creating as many 255-sized segments is incorrect. Does anyone have any idea how to properly set the lacing values?
The common situation when the integrity of an MP3 file is not correct, is when the file has been partially uploaded to the server. In this case, the indicated audio duration doesn't correspond to what is really in the MP3 file: we can hear the beginning, but at some point the playing stops and the indicated duration of the audio player is broken.
I tried with libraries like node-ffprobe, but it seems they just read metadata, without making comparison with real audio data in the file. Is there a way to detect efficiently a corrupted or incomplete MP3 file from node.js?
Note: the client uploading MP3 files is a hardware (an audio recorder), uploading files on a FTP server. Not a browser. So I'm not able to upload potentially more useful data from the client.
MP3 files don't normally have a duration. They're just a series of MPEG frames. Sometimes, there is an ID3 tag indicating duration, but not always.
Players can determine duration by choosing one of a few methods:
Decode the entire audio file.This is the slowest method, but if you're going to decode the file anyway, you might as well go this route as it gives you an exact duration.
Read the whole file, skimming through frame headers.You'll have to read the whole file from disk, but you won't have to decode it. Can be slow if I/O is slow, but gives you an exact duration.
Read the first frame's bitrate and estimate duration by file size.Definitely the fastest method, and the one most commonly used by players. Duration is an estimate only, and is reasonably accurate for CBR, but can be wildly inaccurate for VBR.
What I'm getting at is that these files might not actually be broken. They might just be VBR files that your player doesn't know the duration of.
If you're convinced they are broken (such as stopping in the middle of content), then you'll have to figure out how you want to handle it. There are probably only a couple ways to determine this:
Ideally, there's an ID3 tag indicating duration, and you can decode the whole file and determine its real duration to compare.
Usually, that ID3 tag won't exist, so you'll have to check to see if the last frame is complete or not.
Beyond that, you don't really have a good way of knowing if the stream is incomplete, since there is no outer container that actually specifies number of frames to expect.
The expression for calculating the filesize of an mp3 based on duration and encoding (from this answer) is quite simple:
x = length of song in seconds
y = bitrate in kilobits per second
(x * y) / 1024 = filesize (MB)
There is also a javascript implementation for the Web Audio API in another answer on that same question. Perhaps that would be useful in your Node implementation.
mp3diags is some older open source software for fixing mp3s and which was great for batch processing stuff like this. The source is c++ and still available if you're feeling nosy and want to see how some of these features are implemented.
Worth a look since it has some features that might be be useful in your context:
What is MP3 Diags and what does it do?
low quality audio
missing VBR header
missing normalization data
Correcting files that show incorrect song duration
Correcting files in which the player cannot seek correctly
As part of a project I am working on, there is a requirement to concatenate multiple pieces of audio data into one large audio file. The audio files are generated from four sources, and the individual files are stored in a Google Cloud storage bucket. Each file is an mp3 file, and it is easy to verify that each individual file is generating correctly (individually, I can play them, edit them in my favourite software, etc.).
To merge the audio files together, a nodejs server loads the files from the Google Cloud storage as an array buffer using an axios POST request. From there, it puts each array buffer into a node Buffer using Buffer.from(), so now we have an array of Buffer objects. Then it uses Buffer.concat() to concatenate the Buffer objects into one big Buffer, which we then convert to Base64 data and send to the client server.
This is cool, but the issue arises when concatenating audio generated from different sources. The 4 sources I mentioned above are Text to Speech software platforms, such as Google Cloud Voice and Amazon Polly. Specifically, we have files from Google Cloud Voice, Amazon Polly, IBM Watson, and Microsoft Azure Text to Speech. Essentially just five text to speech solutions. Again, all individual files work, but when concatenating them together via this method there are some interesting effects.
When the sound files are concatenated, seemingly depending on which platform they originate from, the sound data either will or will not be included in the final sound file. Below is a 'compatibility' table based on my testing:
|------------|--------|--------|-----------|-----|
| Platform / | Google | Amazon | Microsoft | IBM |
|------------|--------|--------|-----------|-----|
| Google | Yes | No | No | No |
|------------|--------|--------|-----------|-----|
| Amazon | | No | No | Yes |
|------------|--------|--------|-----------|-----|
| Microsoft | | | Yes | No |
|------------|--------|--------|-----------|-----|
| IBM | | | | Yes |
|------------|--------|--------|-----------|-----|
The effect is as follows: When I play the large output file, it will always start playing the first sound file included. From there, if the next sound file is compatible, it is heard, otherwise it is skipped entirely (no empty sound or anything). If it was skipped, the 'length' of that file (for example 10s long audio file) is included at the end of the generated output sound file. However, the moment that my audio player hits the point where the last 'compatible' audio has played, it immediately skips to the end.
As a scenario:
Input:
sound1.mp3 (3s) -> Google
sound2.mp3 (5s) -> Amazon
sound3.mp3 (7s)-> Google
sound4.mp3 (11s) -> IBM
Output:
output.mp3 (26s) -> first 10s is sound1 and sound3, last 16s is skipped.
In this case, the output sound file would be 26s seconds long. For the first 10 seconds, you would hear the sound1.mp3 and sound3.mp3 played back to back. Then at 10s (at least playing this mp3 file in firefox) the player immediately skips to the end at 26s.
My question is: Does anyone have any ideas why sometimes I can concatenate audio data in this way, and other times I cannot? And how come there is this 'missing' data included at the end of the output file? Shouldn't concatenating the binary data work in all cases if it works for some cases, as all the files have mp3 encoding? If I am wrong please let me know what I can do to successfully concatenate any mp3 files :)
I can provide my nodeJS backend code, but the process and methods used are described above.
Thanks for reading?
Potential Sources of Problems
Sample Rate
44.1 kHz is often used for music, as it's what is used on CD audio. 48 kHz is usually used for video, as it's what was used on DVDs. Both of those sample rates are much higher than is required for speech, so it's likely that your various text-to-speech providers are outputting something different. 22.05 kHz (half of 44.1 kHz) is common, and 11.025 kHz is out there too.
While each frame specifies its own sample rate, making it possible to generate a stream with varying sample rates, I've never seen a decoder attempt to switch sample rates mid-stream. I suspect that the decoder is skipping these frames, or maybe even skipping over an arbitrary block until it gets consistent data again.
Use something like FFmpeg (or FFprobe) to figure out what the sample rates of your files are:
ffmpeg -i sound2.mp3
You'll get an output like this:
Duration: 00:13:50.22, start: 0.011995, bitrate: 192 kb/s
Stream #0:0: Audio: mp3, 44100 Hz, stereo, fltp, 192 kb/s
In this example, 44.1 kHz is the sample rate.
Channel Count
I'd expect your voice MP3s to be in mono, but it wouldn't hurt to check to be sure. As with above, check the output of FFmpeg. In my example above, it says stereo.
As with sample rate, technically each frame could specify its own channel count but I don't know of any player that will pull off switching channel count mid-stream. Therefore, if you're concatenating, you need to make sure all the channel counts are the same.
ID3 Tags
It's common for there to be ID3 metadata at the beginning (ID3v2) and/or end (ID3v1) of the file. It's less expected to have this data mid-stream. You would want to make sure this metadata is all stripped out before concatenating.
MP3 Bit Reservoir
MP3 frames don't necessarily stand alone. If you have a constant bitrate stream, the encoder may still use less data to encode one frame, and more data to encode another. When this happens, some frames contain data for other frames. That way, frames that could benefit from the extra bandwidth can get it while still fitting the whole stream within a constant bitrate. This is the "bit reservoir".
If you cut a stream and splice in another stream, you may split up a frame and its dependent frames. This typically causes an audio glitch, but may also cause the decoder to skip ahead. Some badly behaving decoders will just stop playing altogether. In your example, you're not cutting anything so this probably isn't the source of your trouble... but I mention it here because it's definitely relevant to the way you're working these streams.
See also: http://wiki.hydrogenaud.io/index.php?title=Bit_reservoir
Solutions
Pick a "normal" format, resample and rencode non-conforming files
If most of your sources are all the exact same format and only one or two outstanding, you could convert the non-conforming file. From there, strip ID3 tags from everything and concatenate away.
To do the conversion, I'd recommend kicking it over to FFmpeg as a child process.
child_process.spawn('ffmpeg' [
// Input
'-i', inputFile, // Use '-' to write to STDIN instead
// Set sample rate
'-ar', '44100',
// Set audio channel count
'-ac', '1',
// Audio bitrate... try to match others, but not as critical
'-b:a', '64k',
// Ensure we output an MP3
'-f', 'mp3',
// Output
outputFile // As with input, use '-' to write to STDOUT
]);
Best Solution: Let FFmpeg (or similar) do the work for you
The simplest, most robust solution to all of this is to let FFmpeg build a brand new stream for you. This will cause your audio files to be decoded to PCM, and a new stream made. You can add parameters to resample those inputs, and modify channel counts if needed. Then output one stream. Use the concat filter.
This way, you can accept audio files of any type, you don't have to write the code to hack those streams together, and once setup you won't have to worry about it.
The only downside is that it will require a re-encoding of everything, meaning another generation of quality lost. This would be required for any non-conforming files anyway, and it's just speech, so I wouldn't give it a second thought.
#Brad's answer was the solution! The first solution he suggested worked. It took some messing around getting FFMpeg to work correctly, but in the end using the fluent-ffmpeg library worked.
Each file in my case was stored on Google Cloud Storage, and not on the server's hard drive. This posed some problems for FFmpeg, as it requires file paths to have multiple files, or an input stream (but only one is supported, as there is only one STDIN).
One solution is to put the files on the hard drive temporarily, but this would not work for our use case as we may have a lot of use in this function and the hard drive adds latency.
So, instead we did as suggested and loaded each file into ffmpeg to convert it into a standardized format. This was a bit tricky, but in the end requesting each file as a stream, using that stream as an input for ffmpeg, then using fluent-ffmpeg's pipe() method (which returns a stream) as output worked.
We then bound an event listener to the 'data' event for this pipe, and pushed the data to an array (bufs.push(data)), and on stream 'end' we concatenated this array using Buffer.concat(bufs), followed by a promise resolve.
Then once all requests promises were resolved, we could be sure ffmpeg had processed each file, and then those buffers were concatenated in the required groups as before using Buffer.concat(), converted to base64 data, and sent to the client.
This works great, and now it seems to be able to handle every combination of files/sources I can throw at it!
In conclusion:
The answer to the question was that the mp3 data must have been encoded differently (different channels, sample rates, etc.), and loading it through ffmpeg and outputing it in a 'unified' way made the mp3 data compatible.
The solution was to process each file in ffmpeg separately, pipe the ffmpeg output into a buffer, then concatenate the buffers.
Thanks #Brad for your suggestions and detailed answer!
I was given an uncompressed .wav audio file (360 mb) which seems to be broken. The file was recorded using a small usb recorder (I don't have more information about the recorder at this moment). It was unreadable by any player and I've tried GSpot (https://www.headbands.com/gspot/) to detect whether it was perhaps of a different format than wav but to no avail. The file is big, which hints at it being in some uncompressed format. It misses the RIFF-WAVE characters at the start of the file though, which can be an indication this is some other format or perhaps (more likely in this case) the header is missing.
I've tried converting the bytes of the file directly to audio and this creates a VERY noisy audio file, though voices could be made out and I was able to determine the sample rate was probably 22050hz (given a sample size of 8-bits) and a file length of about 4 hours and 45 minutes. Running it through some filters in Audition resulted in a file that was understandable in some places, but still way too noisy in others.
Next I tried running the data through some java code that produces an image out of the bytes, and it showed me lots of noise, but also 3 byte separations every 1024 bytes. First a byte close to either 0 or 255 (but not 100%), then a byte representing a number distributed somewhere around 25 (but with some variation), and then a 00000000 (always, 100%). The first 'chunk header' (as I suppose these are) is located at 513 bytes into the file, again close to a 2-power, like the chunk size. Seems a bit too perfect for coincidence, so I'm mentioning it as it could be important. https://imgur.com/a/sgZ0JFS, the first image shows a 1024x1024 image showing the first 1mb of the file (row-wise) and the second image shows the distribution of the 3 'chunk header' bytes.
Next to these headers, the file also has areas that clearly show structure, almost wave-like structures. I suppose this is the actual audio I'm after, but it's riddled with noise: https://imgur.com/a/sgZ0JFS, third image, showing a region of the file with audio structures.
I also created a histogram for the entire file (ignoring the 3-byte 'chunk headers'): https://imgur.com/a/sgZ0JFS, fourth image. I've flipped the lower half of the range as I think audio data should be centered around some mean value, but correct me if I'm wrong. Maybe the non-symmetric nature of the histogram has something to do with signed/unsigned data or two's-complement. Perhaps the data representation is in 8-bit floats or something similar, I don't know.
I've ran into a wall now. I have no idea what else I can try. Is there anyone out there that sees something I missed. Perhaps someone can give me some pointers what else to try. I would really like to extract the audio data out of this file, as it contains some important information.
Sorry for the bother. I've been able to track down the owner of the voice recorder and had him record me a minute of audio with it and send me that file. I was able to determine the audio was IMA 4-bit ADPCM encoded, 16-bit audio at 48000hz. Looking at the structure of the file I realized simple placing the header of the good file in front of the data of the bad file should be possible, and lo and behold I had a working file again :)
I'm still very much interested how that ADPCM works and if I can write my own decoder, but that's for another day when I'm strolling on wikipedia again. Have a great day everyone!
Basically I'm trying to replicate YouTube's ability to begin video playback from any part of hosted movie. So if you have a 60 minute video, a user could skip straight to the 30 minute mark without streaming the first 30 minutes of video. Does anyone have an idea how YouTube accomplishes this?
Well the player opens the HTTP resource like normal. When you hit the seek bar, the player requests a different portion of the file.
It passes a header like this:
RANGE: bytes-unit = 10001\n\n
and the server serves the resource from that byte range. Depending on the codec it will need to read until it gets to a sync frame to begin playback
Video is a series of frames, played at a frame rate. That said, there are some rules about the order of what frames can be decoded.
Essentially, you have reference frames (called I-Frames) and you have modification frames (class P-Frames and B-Frames)... It is generally true that a properly configured decoder will be able to join a stream on any I-Frame (that is, start decoding), but not on P and B frames... So, when the user drags the slider, you're going to need to find the closest I frame and decode that...
This may of course be hidden under the hood of Flash for you, but that is what it will be doing...
I don't know how YouTube does it, but if you're looking to replicate the functionality, check out Annodex. It's an open standard that is based on Ogg Theora, but with an extra XML metadata stream.
Annodex allows you to have links to named sections within the video or temporal URIs to specific times in the video. Using libannodex, the server can seek to the relevant part of the video and start serving it from there.
If I were to guess, it would be some sort of selective data retrieval, like the Range header in HTTP. that might even be what they use. You can find more about it here.