Forums » Bugs & Problems Search

Hi,

I'm trying to upload some music for analysis using the Echo Nest Python API, but I'm getting this error:

echonest.web.util.EchoNestAPIError: Echo Nest API Error 8: Computation error (TRISTAN-1)

The error seems to be isolated to certain mp3s on my computer, specifically tracks by composer Arvo Pärt. I've tested a few other mp3s that seem to be analyzed just fine.

Is there any insight you can shed on this issue? I'd be happy to provide any other information if it would be useful for tracking down the problem.

Thanks! Chris

Thanks Chris. The current version of analyze does seem to have trouble with certain audio. We are preparing to release a new version that should fix these problems, hopefully within a week. If you'd like to send me the ID or md5 of the track that is giving you trouble, I can make sure that we run it through the new analyzer to make sure we don't have any problems.

Great, thanks!

The affected MD5s are

a24be7ca85147a2f8c8ccc4fb65674d8 b6e56d4ea356bf506440d036e25a44f3

Meant to have those on two lines for legibility:

a24be7ca85147a2f8c8ccc4fb65674d8

b6e56d4ea356bf506440d036e25a44f3

Hi;

I get a Computation Error 12 when trying to analyse a file. The file is the one with MD5 6c8b883cf9660abb55400834753fbe04 .

I.e., the following link (with key substituted) will trigger it: http://developer.echonest.com/api/get_tatums?api_key=(insert key here)&md5=6c8b883cf9660abb55400834753fbe04&version=3

Adam sent me, btw; he thinks that's probably a bug in the backend.

acb -

Indeed, it looks like we had a problem analyzing that file. It happens on a rare occaision. We'll take a look at it. Thanks.

Paul

I tried several other files and got the same error. They are:

970880adf8c2947c7be02466be8ef522 1fcd4a9f9d6fd92a778cbef298418748

The one thing they have in common is that they're acapella vocals. Could there be an issue with vocals?

acb - how long are the songs? - we also don't like particularly short tracks

IIRC, these were unmixed song "stems", which have long sections of silence.

I thought it would be good to bring up here, b/c in the early mashery-mediated EN Analyze 1.0 days, I tried uploading unmixed tracks like this and found they caused errors much more often than normal, mixed sources. Because of this, I've subconsciously avoided uploading stems since, but it's clear there are some users who would like doing that.

A couple more Echo Nest API Error 12: Computation error (0) I've encountered (with songs between 6 and 12 minutes long) :

d0212de3da1a05ebd0504046d4cb92c1
97301770fe7057d1282521c40c1e2940
4861d8b5d51fca1f2452f3f23208fe38

And the same tracks without their ID3tags (I naively thought that the "metadata" in the traceback could be related to the mp3's meta, so I removed it from the files)

1a0f2d6fbd32e31dfc54b85153d1b8e6
ac704320b8ba9d097511f1f7ee761a80
1203f6c97d7f6c55df45599bfa61117b

Here is the traceback :

Traceback (most recent call last):
  File "C:\_Coder\Python26\echo-nest-remix-examples\reverse\reverse.py", line 48, in <module>
    main(toReverse, inputFilename, outputFilename)
  File "C:\_Coder\Python26\echo-nest-remix-examples\reverse\reverse.py", line 24, in main
    audioFile = audio.LocalAudioFile(inputFilename)
  File "C:\_Coder\Python26\lib\site-packages\echonest\audio.py", line 801, in __init__
    tempanalysis.metadata
  File "C:\_Coder\Python26\lib\site-packages\pyechonest\decorators.py", line 15, in __call__
    self.cache[args] = value = self.func(*args)
  File "C:\_Coder\Python26\lib\site-packages\pyechonest\track.py", line 133, in metadata
    tree = util.call('get_metadata', params)
  File "C:\_Coder\Python26\lib\site-packages\pyechonest\util.py", line 50, in call
    return check_status(response)
  File "C:\_Coder\Python26\lib\site-packages\pyechonest\util.py", line 66, in check_status
    raise EchoNestAPIError(code, message)
pyechonest.util.EchoNestAPIError: Echo Nest API Error 12: Computation error (0)

More :

Computation error (0) 05576a4ea9f17db05bcf90a47e58d20e a86a792e3540ac8cbb90d0bd4cb0a85c

Computation error (TRISTAN-0) 137528216565756f822ab3aca380e043

yhancik - thanks much for the error reports. This is a good opportunity for me to update you on the status of fixing these errors. The Analyzer team has been working very hard on the next version of the analyzer. One of the main goals for this analyzer is stability - it won't crash. Early stability tests are very promising - we've run thousands of tracks through the new analyzer, including hundreds that were difficult for the current analyzer to process. I can't give you a date on when the new analyzer will be released just yet, but I wanted to let you know that relief is coming.

Paul

I've been getting a "Computation Error (12)" also, just for 1 track: "Echoes" by Pink Floyd.
TRBCHNJ1251AE42CB9

Is it possible that the problem is caused by tracks which contain a lot of parts that are quiet or silent or spacey, tripped-out and psychedelic?

I'm getting a "Echo Nest API Error 12: Computation error (0)" too on this hash: 972a5d4910fe32baf1529a99d80c2405

And again for this hash of a different file, same song ("Video Phone" by Beyonce): b5ffc41caea44fe28f44702982b9aeba

The call that is providing the error is get_segments, i.e. http://developer.echonest.com/api/get_segments?api_key=XXX&analysis_version=1&version=3&md5=b5ffc41caea44fe28f44702982b9aeba

Oops, I figured out my error: I was uploading MP4's (videos) instead of MP3's to the API for analysis and getting successful(-ish) responses. It works now with just audio.

cce - I see that you are using analysis_version=1, you should think about switching to analysis_version=3 which is far more stable and in some cases gives better results too.

Paul

I'm having this problem as well. I get a "Computation error (0)" whenever I try to analyze a certain file.

The MD5 is: 145339f6d76d8f029b5fe55f0c029e50

The file in question is quite long -- it consists of an album's worth of songs strung together. Could the length be the problem? I've tried re-compiling and re-exporting the file in different ways and always come up with the same error. I don't really know where to go from here, so any help or insight would be appreciated. Thanks.

ekl, files over 1 hour in length are not currently supported using v3. You might try using v1.

Cheers,

Jason

Thanks for your reply, Jason.

I actually just checked the length of the file, and it's "only" 53 minutes long, so technically, it's under an hour. It's still kind of ridiculously long though. How would I switch to using v1?

Maybe I'm going about this the wrong way. What I'm trying to do is run afromb.py using an entire album as sample material. Is there any way to do that aside from compiling all the mp3s into a single track?

ekl,

You can call v1 by specifying version='1' in your call to analyze: http://developer.echonest.com/docs/method/analyze/

I think what you're doing is the easiest way for you to do afromb without writing code.

cheers,

Jason