Hey guys I know this may either be really awesome... or really lame.. but at least I found it refreshing. This website was on digg and I liked it a lot, be sure to check it out.

http://www.crayonroom.com/moody.php

also check out this short video about how it works, and how the GUI is implemented

http://www.crayonroom.com/moody_quicktour.mov



Basically I was wondering, with the ability of masstagger, and scripts, we could create a mood set of tags and then if someone has the know-how, to add a slider or panel like they have (in this case, a grid) that can be used to create a playlist or as they claim, the best shuffle experience possible. I think this could be a great way to enhance playback and tagging, as compared to the rating system.

I don't know about sliders and such, but i think this can be implemented using existing plugins.

First, we need to make plenty of masstagger scripts (16), one for each "color" in that picture.

Then we link each script to a button. if we do this in column ui, it'll be quite "ugly" but if we do it in panel_ui, i'm sure we can get a grid.

This requires a special plugin to be user-friendly (however, the plugin could be useful for other tasks as well, not just mood - any metadata which has an integer as value).

Creating such a plugin is the easy part. Coming up with a universal mood-system, which is based on psychology and semantics is the hard part. I doubt that people here will be able to come up with a good scheme from scratch.... IMHO it would be a better idea to start off with a plugin in which people can define tags and their value-range themselves.... then let some chaotic experimentation happen..... and then over the course of time learn from what works and what doesnt.

However, for this to be useful, two plugins are actually necessary - one for rating, and another one for creating playlists based on those tags, with just a few mouseclicks.

- Lyx

Interesting idea.
I think a tag cloud plugin would be nice as well a la del.icio.us (instead of just colors).

Should'nt be this very easy?
What is the difference from a rating sytem with stars which is also clickable? That it has two dimensions?
In the posted links I saw in my opinion buttons which createt a special tag....

The only thing which is difficult seems to be a slider....
I which someone would create a stable version of a customizable slider...

In case someone implements a plugin which allows to set customizable tagvalues with a slider - i would be willing to give the mood-system concept a headstart, since i have been working since a few years on a psychological model for mood, behaviour and perception. I would not be able to disclose the model itself, since it isnt published yet - but i could use my findings to create a mood-derivative of it which could serve as the base for mood-based musicrating.

- Lyx

science meets foobar, foobar meets science

but this is cool that it isnt a failure concept. I dont rate songs or albums because that sounds like a lot of work to think 'gee, is this a 3 or 4?' while the video really made sense. It isnt mood per se,.. but the song itself that makes the tag work. In his tutorial, it says that he tags the song based on whether it is a happy song, sad, fast, slow, how light or heavy it seems. All in a two dimensional grid that is in an external application. We just need to find a way to eliminate the external app part, and do this in foobar so that it becomes a second nature thing to tag, like when you are watching your player, you also have the inclination to tag using this grid or slider, compared to thinking about a star rating. The better the GUI involved, the more likely people would use such a 'mood based' or 'perception' based system. Thanks for all the input and I hope this comes together because I dont know squat about the technical side of it all, unfortunately.

Like i said, this is already possible in foobar. the difficulty seems to be in how to judge what "mood" a song falls into.

where does the judgement of mood or color have any 'difficulty'? I dont see where that conclusion comes from, although I may be missing something vital - the user simply choses the color that they associate with a song and the tag is added with that position on the grid. Then further useage comes from creating playlists with that specific mood 'color' and shuffling the colors so that songs randomly play from say ... sad and then move toward lighter and faster beats.




I am interested in that, but dont know what it would do specifically..

this is a prominent feature on many weblogs that indicate a word that is popular based on users who search that item. words that are popular get larger and the smaller ones indicate less browsed items.

but could you give us a bit of an idea how this would work, are there any tags involved? what 'words' would show up in the cloud? artists? albums?

I already have rated and tagged nearly all of my around 4500 tracks with a "tempo" tag. I made masstagger scripts that tag them with a tempo (hard/soft) value between 1 and 5 and added shortcuts. Everytime I listen to a unrated track I tag it (which is not so mush work as I thought).
And I use different colors for hard and soft tracks in my playlist and high rated tracks are displayed lighter. So it's very easy to see "how a track is". I think about adding some mood tags, too. Then I could get nice autoplaylists, but I had to tag all tracks again...

I would be interrested in a system where each user would submit its rating, tempo, mood and maybe other characteristics values for the tracks, too, and I could tag my files with the average values.

It is "possible" in a very hackish patchwork-style way - nothing even remotely clean and easy to setup and use. Even though, making things possible in a hackish way seems to be the current trend, i do not support this attitude.

- Lyx

our own little freedb world.

but Birk could you show us what your script is like, or how you impliment colors more specifically? the tempo is based on an actual beats per minute or data that is part of the song correct? it isnt the same as happy/sad, but it seems more like hard/soft based on an actual calculated value. But this is in the right direction I think.

the last part of your post suggested that our tags get stored somewhere that we could compare them to each others? I dont know if that is possible since we would need to create a database. This topic is actually happening right now with the topic about music genome project. I am all for it, but there are cost and technical aspects that make it difficult to do on a scale that we are imagining.

I'm watching this video again on the moody website and this dude sounds soooo slick. Haha but I like the idea of how the color swatches have tabs at the top : one for 'listen' and one for 'tag' which is good.

there is a feature called quicktag that he uses which allows the playlist to continue until you tag it with a color. This allows for really fast tagging of a large collection.

can we do either of these in a new plugin?

About tagging tracks, I created one masstagger script for each value of my "rating", "tempo" and "situation" tag and through the use of keyboard shortcuts, Girder and a Remote control I tag them while listening to them.

About a central repository for people taggings, I'm a bit scepctic because it happens to me to sometime "disagree" with the tagging I put in a track days before.

About a mood selector, I choose "situation" instead of "mood" because the second is... too moody while for the first I just have to imagine (judge) which situation a track would fit.
I have set these "situations" with some [characteristics] and (examples):

- breakfast = [any tempo, clear melody lines and easy to follow variations...] (most of Bach, many of Mozart, Scarlatti, Satie, Strauss, Children's Album Leaves etc)
- house cleaning = [fast, energic, little volume variations...] (Some "Beethoven, most of fast rock and pop, most of be-bop etc)
- hugging times (we calle it "moony") = [medium to slow, lovely melodic lines, crooning, close miking - instruments or voice...] (Tuck and Patty's "My romance", Ella Fitzgerald's "In a sentimental mood",Lester Young, some Miles Davis, some Tom Waits, a lot of pop)
- conversation = [any tempo, not too energic, easy to follow, well known songs, little volume variations] (mainstream jazz, Tin Pan Alley, traditional pop, hard bop, Ella Fitzgerald, Franck Sinatra, Louis Armstrong, Paolo Conte, Sting, Capossela...)
- listening = [not easy to follow, wide volume variations, ugly but historically unmissable] (some Miles ('60s mostly), many classical with "pp" to "ff", most of Ornette Coleman, some Coltrane...)
- relax = [like "moony" but absolutely without a loudness spike, not even that lovely trumpet solo attack" - you would hate it while listening in bed with a headphone trying to chatch Mr Morpheus...] (Sarah Vaughan, Paul Desmond & Jim Hall, some crooning...)

About moods, I found some models developed from different people in universities and I think it's quite complicated trying to define a mood scientifically. For instance, one model says you should judge how much "energy" there is the music but the same level of energy could end up in different moods according also to the level of "pleasure/displeasure" you get from it. Somebody did even say that it could be possible to "auto find" the mood of a piece of music with a well thought out software (measure insted of judge). Mah?!?!

I have solved that.


This is a stupid idea because it will not work without cloning human consciousness. We are not just talking about energy, speed, etc - there are factors at work on the conceptual and semantic level - basically what the music "means" - instruments cannot measure that - you need a consciousness AND conceptual thinking for that.

- Lyx

Lyx,

you are not making the slightest bit of sense on thisissue. My 'stupidest' idea has nothing to do with human conciousness. your comments about meaning are not involved. Folks, please,
let me explain exactly what the mood tagger, and playlist idea is about...

1. you have scripts that assign a number, or name/color to the song. 1-16, or however many moods there are in our plugin.

2. To tag the files, instead of using stars, or manually entering in something, we pick them from a nifty color grid box exactly as the video shows.

3. once the tags build over the collection, you can create playlists that only play those 'moods' or create a autoplaylist that takes moods from say... 5-16 and plays them in increasing order (suggesting that 5 is dark and 16 is light) so your music is going to higher tag values.

4. That is IT, nothing more, foobar is not guessing your mood or learning anything, or making you a sandwich. Certainly there is no 'energy, speed, etc' no conceptual or semantic level stuff(s).

these tags are simple and replace the 1-5 stars rating. I cannot imagine how this is any more complicated.

Apparently, you have proposed something of which you have no clue how it works. Weeeh, i love people who blow lots of hot air without knowing what they are talking about.

"I dont understand it, so you must be wrong" yesssssss! Ignorance rules!

- Lyx

Peace, people,

I think both positions have their merit.

LyX wants some advanced algorithm that can relate mood tags to some other measurable variables. He may
well have a model on how these measurable variables relate to what we call 'mood'. This could be tremendously useful for tagging music, mainly because, if I understand it right, it could be automatic. Building a machine-learning algorithm on it (something that may resemble what musicIP does, but produces mood tags) could be extremely cool. Again, I don't know anything about LyX work, so I'm just guessing here.

Rama wants an UI to be able to tag *manually* a music collection as easy as possible. This is very interesting too because obviously people may want to tag music with all kind of personal tags (i.e., 'music for fishing' ). No algorithm can give you that level of personalized tagging -unless you want to code your own-.

Ideally, we could have both.

Again, correct me if I'm missing your point.

I created a new thread for discussing mood-systems in general, since according to the threadstarter, this thread is only about the crayonroom system. Thus, discussing other concepts is offtopic here and should instead be done in the following thread:

http://www.hydrogenaudio.org/forums/index....showtopic=55038

Is that supposed to be an insult because Moody relies on gut feeling instead of deep science? I actually think that music, colors and gut feeling make a pretty good combination. I'm not saying you should abandon all other approaches, but remember that some people may want to use a mood based tagging system without writing (or reading) a thesis about it.

From what I see on the website and in the demo video, Moody is extremely simple at its core: Moody allows you to tag each song with on of a set of discrete values. Moody also allows you to pick a number of those values and it will play only songs that are tagged with one of them. That alone isn't so amazing that you would fall of your chair.

What sets Moody apart is the presentation. Instead of using numbers or text labels for the values, it uses colors. The colors are arranged in a two-dimensional grid with labels on the axes to provide a guideline for the user. There are no strict rules for tagging songs. If in doubt, users have to turn to their gut feeling. In my opinion, the most important reason for providing guidelines at all is that the mood colors/tags are useful for other people, for example for a friend who picks out music on your computer.


On a side note, I know there is an audio player that lets you tag songs with colors without providing any rules or guidelines how that should be used. The colors come from a cyclic band and are used to build smart playlists with soft color transitions. Unfortunately, I remember next to nothing about that player. I'm sure it was for Windows and I think the color tagging was part of its core features. If anyone knows which software I'm talking about, please drop me a PM.

Just to point out, as no-one seems to have specifically stated this, that it appears from the video that Moody simply sets a comment tag, using the format:

Moody<Row (A-D)><Column (1-4)>

... using the matrix:


E.g.: "MoodyC2" or "MoodyB3".

In its simplest form, you could then use Autoplaylist manager to create a playlist using "%comment% IS MoodyA4".

I have no idea about panels, etc. - I don't tend to pay too much heed to foobar's appearance - but I guess you really need a developer to create a pretty panel that lets you click on a square to essentially do the same thing.

I don't know if Moody does this, but it seems to me that it would be nice if it weighted the values somehow, so if you clicked on C2 you may also get B1, B2, B3, C1, C3, D1, D2, and D3 tracks, but in a lot smaller supply than actual C2 tracks.

I did say what i did say. I did not say what i did not say.

Thank you foosion, it is really simple at its core, as I have stated.

This was truly a simple exercise in creating a panel that does the same job all rating systems have been doing for years (albeit, just simplified to a color grid).

I endorse all discussion on topics and generally open to most things - however there is no need for derisive or evasive argument. Please, we all are here for the same reason -music and a great program foobar, not to prove anyone is better or worse. Lets keep it that way.

That itself should be simple to build (relatively). Implement it as a side panel, click the color, sets the tag. I guess I haven't done any programming with foobar, so... well, yeah.

It seems like you'd have repeatability problems with this model--theres 16 different choices, on different days people might choose another one of the 16 colors. Just choosing the neighboring colors would give you 3/4 of the tracks back.

I think making this more complex would might make things more interesting. Adding an aspect of community consensus (this would require some online database) might lead to some more interesting picks for mood playlists. Add more than 16 choices, maybe a 32x32 (or bigger) plane to choose from. Mood playlists could be created with tracks tagged in a radius from an initial point, a community consensus point, or the midpoint of the user point and community point. I'm rambling. It's late.