Picking wrong format as "best"

[sandebert]

Hi,

I think I have found an edge case where youtube-dl doesn't detect the highest quality format. It should be noted that this is a very unusual video there the highest resolution is 2160p.

I'm using youtube-dl version 2015.01.07.2.

I realise there are more reported issues on this topic, but I'm still putting in a report as I hope this provides extra information that might be valuable. (As in, it's not really a judgment case as whether do download a webm or mp4 when they are identical sizes, as reported in #3482.)

When I list the formats using

youtube-dl -F --verbose "https://www.youtube.com/watch?v=udAL48P5NJU"

I get this

[debug] System config: []
[debug] User config: []
[debug] Command-line args: ['-F', '--verbose', 'https://www.youtube.com/watch?v=udAL48P5NJU']
[debug] Encodings: locale UTF-8, fs UTF-8, out UTF-8, pref UTF-8
[debug] youtube-dl version 2015.01.07.2
[debug] Python version 2.7.3 - Linux-3.2.0-4-amd64-x86_64-with-debian-7.7
[debug] exe versions: avconv 0.8.16-6, avprobe 0.8.16-6, ffmpeg 0.8.16-6, ffprobe 0.8.16-6, rtmpdump 2.4
[debug] Proxy map: {}
[youtube] udAL48P5NJU: Downloading webpage
[youtube] udAL48P5NJU: Extracting video information
[youtube] udAL48P5NJU: Downloading DASH manifest
[info] Available formats for udAL48P5NJU:
format code extension resolution  note 
139         m4a       audio only  DASH audio   48k , audio@ 48k (22050Hz), 1.20MiB (worst)
171         webm      audio only  DASH audio  113k , audio@128k (44100Hz), 2.57MiB
140         m4a       audio only  DASH audio  129k , audio@128k (44100Hz), 3.19MiB
172         webm      audio only  DASH audio  167k , audio@256k (44100Hz), 3.75MiB
141         m4a       audio only  DASH audio  255k , audio@256k (44100Hz), 6.33MiB
278         webm      256x144     DASH video  138k , webm container, VP9, 15fps, video only, 2.09MiB
160         mp4       256x144     DASH video  111k , 15fps, video only, 2.60MiB
242         webm      426x240     DASH video  186k , 30fps, video only, 2.89MiB
133         mp4       426x240     DASH video  265k , 30fps, video only, 5.99MiB
243         webm      640x360     DASH video  310k , 30fps, video only, 5.67MiB
134         mp4       640x360     DASH video  605k , 30fps, video only, 11.92MiB
244         webm      854x480     DASH video  590k , 30fps, video only, 10.77MiB
135         mp4       854x480     DASH video 1129k , 30fps, video only, 24.88MiB
247         webm      1280x720    DASH video 2128k , 30fps, video only, 26.87MiB
136         mp4       1280x720    DASH video 2234k , 30fps, video only, 50.17MiB
248         webm      1920x1080   DASH video 6555k , 30fps, video only, 57.81MiB
137         mp4       1920x1080   DASH video 4407k , 30fps, video only, 94.93MiB
271         webm      2560x1440   DASH video 57251k , 30fps, video only, 549.30MiB
264         mp4       2560x1440   DASH video 10047k , 30fps, video only, 223.53MiB
313         webm      3840x2160   DASH video 161626k , VP9, 30fps, video only, 1.32GiB
272         webm      3840x2160   DASH video 161626k , 30fps, video only, 1.33GiB
266         mp4       3840x2160   DASH video 22306k , h264, 30fps, video only, 487.68MiB
17          3gp       176x144     
36          3gp       320x240     
5           flv       400x240     
43          webm      640x360     
18          mp4       640x360     
22          mp4       1280x720    (best)

I suspect that formats 272 or 313 probably should be selected as best.

[phihag]

As indicated by the output, those formats are video only, i.e. do not contain audio. Most of our users (including myself) prefer a 720p video with audio to a 2160p video without.

You can use -f bestvideo if you are fine with videos without sound, or -f bestvideo+bestaudio to download the best video and the best audio format and merge them together. Since this requires external software (ffmpeg/avconv) and results in a couple of artifacts (for example you won't be able to play the video while it's downloading), this is not yet the default, although it certainly may be one day (patches welcome!).

[sandebert]

Aha, I didn't understand that aspect of "best". Yep, seems perfectly reasonable - thanks for helping out.