[xexe]

After a couple IRC chats with capnbry (of HD Tagging fame) it was decided to start a new thread around the theory and practice of source tagging.

What is source tagging?

Any means that can derive the "source" a video came from. Common sources include, Bluray, HDTV, PDTV etc

So why do we want source tagging?

Rubbish in rubbish out. The source is a good "indicator" of the quality of the final video result.

But I have seen sources tags in skin xxx
Currently some skinners extract a few source tags from the video filenames such as Bluray but we are interested in XBMC core functionality and not skin specific implementations.


The immediately obvious method of classifying source is from the file name. Many people use a source tag in their filenames and in many cases they are very predictable. Using REGEX we could trivially extract these tags and normalize them into a defined set as yet to be scoped.

Sources like HDTV and PDTV are obvious however there is a confusion over how some source tags actually refer to quality as well. For instance the source tag BDRip refers to a source of Bluray but also means encoded to xvid. We only care about deriving the source.

This thread is dedicated to ideas on how best to identify and classify sources in a manner similar to how HD tagging was eventually implemented in the database. Specifically its not about the politics of it, external nfo tools, how skins should or shouldn’t work etc or anything else; just ideas on the theory and practice.

So apart from filenames are there any other methods for determining the source?.
What is the definitive set of sources we should be working with this now?.
Any other ideas?.

[xexe]

Reserved for documenting findings

[HunterPro]

Most important for quality isnt the filename but the bitrate, compression and resolution. Based on those informations you can give a quality rating.

Giving out a 'bluray' logo is always going to be insane since you're not playing bluray. You're playing a bluray rip.

The only sane things you can say is 1080p/i / 720p / SD and stereo/5.1 / lossless.

[ronie]

Very interesting!
I've been giving this a lot of though when i was creating the Transparency! skin. I didn't like the idea of having to rename all my movies to include a source tag, just to show a source flag in the skin.

Since then, i've developed a set of matching rules that will 'guess' the source if no source tag is present in the filename. Some of these rules are 100% accurate, while others....well, are just a wild guess.

Basically what i'm doing is looking at the file-extension.
If the extension is .ifo or .vob, the source is clearly a DVD.
Same goes for the BluRay specific extension, .mt2s files will be flagged as BluRay.
HDDVD's also have their specific extension (.evo) so those are flagged as HDDVD.
These matching rules are the 100% accurate ones.

As for the less accurate ones, feel free to ignore these.... i'm just listing them to hand out some ideas...
1) Movie Library
- For all other extensions (.iso .mkv .avi .mpg etc...) 'assume' the source is BluRay if the VideoResolution is 720 or 1080. The rest will be flagged as DVD.

2) TV Show Library
- Same as above, 'assume' the source is HDTV if the VideoResolution is 720 or 1080. The rest will be flagged as SDTV.


I agree with what HunterPro mentioned above, the source tells you something about the source only, it has little to do with the quality of the rip.

[xexe]

Sounds very reasonable. I dont want to touch much on quality and that why it carefully said that source is an "indicator" only.

The main thing is that it is information that is useful in the library if you are a collector and useful via the gui if you are a user.

In reality I don.t think we have to make a case for its usefulness since it has already been well received in the skins that have implemented so far.

I have started making notes on what inputs relate to what source and will post them in the next couple of days once i get some more thoughts.

[minimalko]

Implementing the source information into XBMC's library sounds useful. However It is really hard to imagine how XBMC should know about the source without reading it from filename.

EDIT:One way would be to create a database of all the titles released on HDDVD (there is not too many of them, and there won't be any more). Perhaps extra field in TMDB would do the job.

In case local movie in the library is at least 720p, but was not released on HDDVD it would be considered Bluray. That of course doesn't solve problems with HD videos that were ripped from other HD source but those could be adjusted manualy. It would still work better than renaming the files IMO.

[watzen]

minimalko: I believe that the plan was to read the filename. Just to do it in the code instead of doing it with visibility conditions, as the visibility conditions take a lot of cpu cycles.

[minimalko]

minimalko: I believe that the plan was to read the filename. Just to do it in the code instead of doing it with visibility conditions, as the visibility conditions take a lot of cpu cycles.

Well thats the obvious method, but would not actually change that much. Renaming the files (+ subtitles in many cases) and rescraping the whole library seems just too much work.At least from my point of view. Plus with all the conditionals you find in skins this change would not really be noticable.

So apart from filenames are there any other methods for determining the source?.

I just tried to answer to this part.

[jmarshall]

There is no way to decide on a source, other than the user specifying it themselves.

All we need to decide, then, is how this recognition is done. I have no problem reading a <source> tag from an xml file. I'd prefer not to scrape the filename for it, but agree that that may be the most useful option.

EDIT: Moved studio related discussion to a new thread

Cheers,
Jonathan

[snoxbox]

Most important for quality isnt the filename but the bitrate, compression and resolution. Based on those informations you can give a quality rating.

The only sane things you can say is 1080p/i / 720p / SD and stereo/5.1 / lossless.

I disagree. We have the media flagging for resolution already implemented.

Giving out a 'bluray' logo is always going to be insane since you're not playing bluray. You're playing a bluray rip.

I have a combination of 1080p mkv's: Some are scene releases, compressed down to ~10gb. Others I have ripped myself from blu-rays, and are uncompressed (~20 - 30gb). No, it's not strictly playing from a blu-ray, but if I was to play it from a blu-ray it woiuld be exactly the same quality, as I have not compressed it. I'd like to see a blu-ray logo for these, and 1080p for the scene releases.

[RockDawg]

One option could be to scrape the name of the source in XBMC. By 'source', I mean when you "add source" in files view. For example, my current source name for my Blu-Ray movie rips is "Blu-Ray Movies". Currently, my DVD movie rips source is simply named "Movies". It would be cake to just change it to "DVD Movies". Much easier for a user to change the name of a few sources rather than hundreds of files. It may not work for everyone, but it would be a good option for a lot of us.

[The-Boxhead]

Okay my toughts...

So there is no way to read file metadata to get this ?

xbmc team has always had a idea that it should work for all user, not only the techy ones. Offcourse the ones who hang around in these forums is mostly computer know how's and know how to fix stuff themself. But the casual user will probably have 50 video files floating on the computer, barely knowing where they all are. They have no idea about what the file names should be, what nfo files is or what not.

To bad that video files doesent have id tag's like mp3 files

Right now there is a mess getting all this tagging working - nfo files and file naming -movietrailer and all that. it's a mess even for techy users sometime. It's floating around 3-4 movie programs that all does stuff a little different.

But im all for a unified system embedded in xbmc itself - the problem seems to be that there is no reliable way of getting this info ?

I really dont care about what studio has created a movie, why would we need that ?
But the quality is important - if its SD - 720p or 1080p
The sound is also important.
I guess for me Sound and Resolution are the main ones.

[jmarshall]

The-BoxHead: XBMC already reads that information.

This thread is discussing the "source" information tagging which is currently being done in some skins by reading the filename.

Obviously real data such as resolution/codec info is read direct from the file already.

[The-Boxhead]

Ahh then thats good enough for me.
Been quite a mess in my movie source lately so got a little
confused about what my mediamanager/skins did as to what xbmc did

[CapnBry]

MODEDIT: Moved studio related discussion to a new thread

Back on track: I can't think of any other way to do it besides either using the source name (which is a great idea, RockDawg) or the filename or maybe even the path name. There's no big list of titles you can check and determine if "Gone With the Wind" on your hard drive is an original CAM release or the 1080millionP version. If it is in an NFO file, then that obviously is the best way to tell.

Maybe something like a standard list of <source>=<substring>? I'm really wide open on how to do this.