Asian characters restricted in filenames.

[seojoohyun]

Using Windows 7. I was having some problems downloading a video so I did a -U to update to the latest version (2015.02.23.1) to see if that would help, but after the update I no longer get the full file names, specifically east asian characters are being stripped out.
For example on a video like this one:

https://www.youtube.com/watch?v=ouR4nn1G9r4

instead of (하이포, 아이유) I just get (, )

Edit.... I usually use a batch file to run youtube-dl, with something like this in it:

youtube-dl.py -w -f22 -o %%(title)s-%%(id)s-720.%%(ext)s http://www.youtube.com/watch?v=%1

[phihag]

Can you post the output you get when you add -v to your command line? That output tells us how to reproduce the problem.

[seojoohyun]

C:\Videos\ytdl>youtube-dl.py -v -w -f22 -o %(title)s-%(id)s-720.%(ext)s http://www.youtube.com/watch?v=ouR4nn1G9r4
[debug] System config: []
[debug] User config: []
[debug] Command-line args: ['-v', '-w', '-f22', '-o', '%(title)s-%(id)s-720.%(ext)s', 'http://www.youtube.com/watch?v=ou
R4nn1G9r4']
[debug] Encodings: locale cp1252, fs mbcs, out cp437, pref cp1252
[debug] youtube-dl version 2015.02.23.1
[debug] Python version 2.7.6 - Windows-7-6.1.7601-SP1
[debug] exe versions: rtmpdump 2.4
[debug] Proxy map: {}
[youtube] ouR4nn1G9r4: Downloading webpage
[youtube] ouR4nn1G9r4: Extracting video information
[youtube] {22} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] {43} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] {18} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] {5} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] {36} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] {17} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] {137} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] {248} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] {136} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] {247} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] {135} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] {244} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] {134} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] {243} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] {133} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] {242} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] {160} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] {140} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] {171} signature length 40.41, html5 player en_US-vflq_9mxe
[youtube] ouR4nn1G9r4: Downloading DASH manifest
[debug] Invoking downloader on u'https://r4---sn-8xgp1vo-2ias.googlevideo.com/videoplayback?ms=au&ip=84.17.35.33&requir
essl=yes&mime=video%2Fmp4&key=yt5&mt=1424822818&mv=m&sparams=dur%2Cgcr%2Cid%2Cinitcwndbps%2Cip%2Cipbits%2Citag%2Cmime%2C
mm%2Cms%2Cmv%2Cpl%2Cratebypass%2Crequiressl%2Csource%2Cupn%2Cexpire&id=o-AEUuKewraoOmi-vUxX4ZH-sexm4aj4ZAGs0bsfTElDtv&pl
=20&mm=31&itag=22&initcwndbps=1657500&source=youtube&upn=sdTnuEMsHvU&gcr=us&sver=3&ipbits=0&fexp=905657%2C907263%2C92762
2%2C934954%2C936118%2C943917%2C947225%2C947240%2C948124%2C951703%2C952302%2C952605%2C952612%2C952901%2C955301%2C957201%2
C959701&dur=204.335&ratebypass=yes&expire=1424844443&signature=726BB420B0CA4B030480055EABEFECF755D612DC.6F0A9BF66AD13FF9
40445C4685E66BC2FE794189'
[download] Destination: [MV] HIGH4, IU(, ) _ Not Spring, Love, or Cherry Blossoms(,, )-ouR4nn1G9r4-720.mp4
[download] 100% of 35.43MiB in 00:09

[dstftw]

That's caused by 3a0d2f5 as workaround for #4787 on python2 @ Windows.

[seojoohyun]

Do I understand this correctly..... in order to be able to pass the filenames to ffmpeg/ffprobe for post processing, the filenames are stripped of these characters when saved?

[dstftw]

@seojoohyun exactly, or replaced with 'counterparts' according to preferredencoding.

[seojoohyun]

I don't have any need for post processing with ffmpeg, but I do want the complete filenames, so I modified my local copy by commenting out those lines added in the workaround. That makes it work for my purposes.
I take it that upgrading to python 3 would eliminate this problem?

It would be nice if there was a command line option to disable it trying to use ffmpeg. For example I don't want the DASH audio files converted to work with itunes or whatever, but it tries to do it automatically and keeps giving me a warning.

[casinero]

Another proposal: adding a command-line switch which offers complete filename like old times, and simply stating in the help docs that this switch might conflict with post-processing feature.
So the overall default behavior of youtube-dl doesn't change, and people who do want both post-processing feature and complete filenames like me can always do the following things manually or scriptingly: normally download videos with whatever post-processing features it takes (actually even --restrict-filenames could be used to prevent similar locale/encoding issues), and use --get-filename or -j combined with the aforementioned switch to get the complete filename, then rename by ourselves.

[Steveland]

Why is youtube-dl.exe still using python version 2? updating to version 3 should solve the issues with unicode.

[dogancelik]

I want to have original filenames when doing -f bestvideo+bestaudio. Is there a solution for this?

[dstftw]

Fix for this issue has just landed in youtube-dl 2016.05.30 along with switch to Python 3 for Windows build.