Unable to play an M4A file #2055
-
|
I'm trying to play files over FTP using a MultiDecoder. It works fine with MP3 and WAV: However M4A despite determining the type correctly I only get boinks and farts, and the decoder seems to be very confused: I have also found the ContainerMP4 class, but using that instead of the AAC Helix codec just gives me silence and the whole system appears frozen. Is there no solution to playing M4A files just yet? Thanks |
Beta Was this translation helpful? Give feedback.
Replies: 9 comments 45 replies
-
|
M4A is a container format: I am not aware of any demuxer implementation that would run on a microcontroller... |
Beta Was this translation helpful? Give feedback.
-
|
Isn't the class I've linked at the end ( ) a barebones demuxer that exists in this very project?ESP-ADF has (somewhat limited) M4A support too, so in theory that is possible: https://github.com/espressif/esp-adf/blob/4200c64d4a78b9e9e2297818733aff21c72079a1/examples/cli/main/console_example.c#L387 (My Nokia phone from 2006 did M4A, and an ESP32 is way more powerful than that, so...) |
Beta Was this translation helpful? Give feedback.
-
|
Would you volunteer to test a new m4a demuxer and eventually a related alac decoder ? |
Beta Was this translation helpful? Give feedback.
-
|
Sure thing, I can do that, although I don't have any ALAC files remaining after migrating from iTunes when they destroyed it in 2017. But I have plenty of AAC/M4A ones! Is that the thing I've linked? If so, I got it to compile, but it didn't work for me, just freezes until the WDT kicks in. Fighting other issues in my own code in the mean time :-) |
Beta Was this translation helpful? Give feedback.
-
|
I committed some DRAFT ContainerM4A class: Just pass your MultiDecoder in the constructor and add an entry for "audio/m4a" which points to the ContainerM4A to the multidecoder. |
Beta Was this translation helpful? Give feedback.
-
|
Just tried with latest master and after fixing a few typos it's still the same freeze with the strange size. It feels like it's missing some header unwrap before trying to go into the box decoding but I really need to dive into the code to find what exactly and I didn't have the time yet in the past days :( |
Beta Was this translation helpful? Give feedback.
-
|
I had a hard time to get the ALAC codec to work, but finaly it started to do what I expected. In order to be able to debug the m4a parsing, I added a test case that can be compiled and debugged on the desktop. |
Beta Was this translation helpful? Give feedback.
-
|
I'll try the latest version later tonight. Do you happen to have any guide about how to test it on desktop? As I have a few other things that are a bit confusing to debug and it seems that enabling I2S is breaking JTAG so I cannot debug on the device now... |
Beta Was this translation helpful? Give feedback.
-
|
I was finally finding why a nested MultiDecoder was not working as expected: it seems that it closes the last decoder when a new decoder was selected: This was causing that it was closing it self! This was easy to fix with an additinal check... |
Beta Was this translation helpful? Give feedback.
-
|
Oh, interesting. |
Beta Was this translation helpful? Give feedback.
-
|
In the example directory that I provided just execute the following This builds an executable that you can use with the debugger. |
Beta Was this translation helpful? Give feedback.
-
|
Tried the latest version, but it's crashing I guess due to (seems that the vector at fault is not created if that STSZ error is logged) If I set the box buffer size to be bigger than the default 8192 to a somewhat ridiculous 65536 bytes, then I can go past this error, so the box is being read correctly probably — but then there is an endless loop in the actual parsing right after that condition. |
Beta Was this translation helpful? Give feedback.
-
|
OK, the loop is not actually endless, it's just super long because of the use of a vector without resizing it in advance. Once it ends it goes further without a crash, albeit still no sound, and more weird boxes: |
Beta Was this translation helpful? Give feedback.
-
|
I was using generating the ContainerM4A implementation with AI. After digging into it I am not sure if I want to keep this! Can you share a m4a file ? |
Beta Was this translation helpful? Give feedback.
-
|
Interesting, I'll check it out! Do you mean the album artwork by the image? |
Beta Was this translation helpful? Give feedback.
-
|
yes |
Beta Was this translation helpful? Give feedback.
-
|
I also added an API to use the Arduino File API: this avoids the need to store any data. |
Beta Was this translation helpful? Give feedback.
-
|
Here are some related blogs |
Beta Was this translation helpful? Give feedback.
-
|
The idea of using Redis for the sample list was interesting. Talk about "Hello world Enterprise Edition" :-) One thing many forget about though when using beefier top of the line ESP32 modules is HIMEM. It's essentially a second PSRAM bank of extra 4 megabytes. It cannot be mapped to regular address space IIRC, but if this list is accessed so rarely that it even works over LAN with Redis or SD card speeds, then it might be a perfect candidate for the task. It requires special functions to access it, but since you have the buffer abstraction already, that might be easy enough. |
Beta Was this translation helpful? Give feedback.
-
|
Oh, that's a really cool idea. I never looked into the non adressable part of PSRAM. |
Beta Was this translation helpful? Give feedback.
-
|
I could try although time has been scarce recently, I feel bad for asking for this and never coming back yet to test how it works :/ The documentation for the HIMEM module is also not very clear to me. It says you map a part of it into the address space, but how does it work exactly it does not say. E.g. does it replace a part of the address space like it was on the zx spectrum? If so there needs to be extra precautions so that an interrupt won't try to read the original PSRAM while it's mapped out. Or does it map it to an unused address block but that is too small to map the whole extra 4M of what... |
Beta Was this translation helpful? Give feedback.
-
|
I also had a quick look at the documentation: I think it is also unclear if this is supported by Arduino, since some config changes are needed. |
Beta Was this translation helpful? Give feedback.
-
|
Seems like people are using it with Arduino: https://github.com/arkhipenko/ESP32Himem Edit: that code is copy pasting the implementation from IDF into its own repo so maybe it's not good as things might break with changes to hardware. Also here's the answer, it cuts off a chunk of the PSRAM address space to access that. So some clever page juggling might be needed in order to use both at once. Which might also imply needing to disable interrupts and such while the read/write process is ongoing, to avoid other code still expecting PSRAM data accessing the wrong thing. |
Beta Was this translation helpful? Give feedback.
-
|
I decided to roll my own library which provides some more Classes to use PSRAM and HIMEM |
Beta Was this translation helpful? Give feedback.
-
|
Went to test today but it won't build with C++17: Replacing with box.id = (this->box.id + 1); seems to work, checking further. Also a lot of stuff seems to be redefining the standard library in Net.h, this is not ideal either but hopefully won't explode things — in particular the ATAPI driver in my project relies on working h2be/be2h |
Beta Was this translation helpful? Give feedback.
-
|
Yes, now I can hear sound! Checking other file types next... |
Beta Was this translation helpful? Give feedback.
-
|
Surprisingly it decodes HE AAC files just fine, too. Even on an iPod Classic those files resulted in a loss of a lot of high frequencies because of the limited CPU power |
Beta Was this translation helpful? Give feedback.
-
|
OK, it was going all too well, but it seems now MP3 playback broke... |
Beta Was this translation helpful? Give feedback.
-
|
Seems like MP3 is detected as MP4 by the MIMEDetector: |
Beta Was this translation helpful? Give feedback.
-
|
That should be easy to fix by checking if the box type characters are all alphanum |
Beta Was this translation helpful? Give feedback.
-
|
I think this is a false positive on MIMEDetector's side. It's looking for an
Probably this is what's causing it |
Beta Was this translation helpful? Give feedback.
-
|
I agree: files starging with id3 are not m3a! |
Beta Was this translation helpful? Give feedback.
-
|
Not sure why this would be reported as MP4. |
Beta Was this translation helpful? Give feedback.
M4A is a container format: I am not aware of any demuxer implementation that would run on a microcontroller...
You will need to convert thses files on the descktop to aac files e.g. using ffmpeg