Mixest was a project I built this summer because Pandora is too mainstream and Spotify isn't random enough. I wrote a scraper and crawler in node.js that constantly searches for new blogs and adds new music to the radio. Instead of an account system, I used HTML5 local storage to keep track of favorite songs and recently played songs.
You may have seen Mixest on HN 3 years ago, back when it was written in PHP. Its data was destroyed last year by a PHP virus, so I rebuilt it from scratch.
I've been using it as my hacking playlist, and I hope you guys will enjoy it too!
I ran a vaguely similar music discovery service called Albumcorner earlier this year. I got in touch with the guy who did Mixest -- seemed like a really great guy. I got the impression that he wasn't going to relaunch after the attack on the server, though, so it's great to hear he passed the baton. And it looks better than ever, Michelle, nice work!
Congrats on re-launching! Could you tell us a little about the tech? I think some folks would be interested in hearing about the node crawler and maybe a little about how it distinguishes indie music from the rest of the content out there.
I am absolutely loving this. Any chance of more permanent accounts though? Local storage/cookies are something I go through quite often, plus I'd love to be able to go back to songs at a friend's house or party and play some of these.
I had emailed you previously when the first incarnation of the site died -- so glad it's making a comeback! (EDIT: Or maybe I had emailed the original authors? Not sure on the timeline I guess.)
Great execution, no barrier to playing.
If I understand what you are doing correctly (scraping mp3 blogs), most of the music is provided to the blogs by artist (or PR) for purposes of promotion, so legally it's relatively solid.
I'm curious to know what blogs you are scraping, and how the crawler is moving on to new blogs? I chase a lot of music blogs and host one myself, and know how inconsistent they can be design and format-wise!
Also suggest linking to the original blog the song was found from, somewhat of a courtesy to the blogger, but even more as a means to find more information about the artist (without you having to input that data yourself)
Great stuff! I've really been enjoying this--the interface is clean and it's a great experience. If you write up more about how you find the songs or the tech under the hood, I'd enjoy reading more.
Also, I'd love to see buy/download/more info links. It would be a lot of work to canonicalize band/song names, but might provide some good ways to monetize if you wanted to make this a bigger project.
The site is incredible. I love the design and the extraordinary simplicity in functionality.
To make a side comment, this seems like an ideal MVP. It gets all the core functionality well-implemented, with enough polish to make it captivating.
Some users mentioned that user accounts, despite the use of HTML5 to store user-specific info, is something that would still be useful. I agree, and it makes it an ideal part of your next revision.
Keep up the good work! This is definitely bookmarked.
Love this, already found a great new artist I love. Well done. Another vote for a little more info on where the track was found so information on the artist can be chased down.
+1 for this. More background info in general (site where you found it, artist(s), album(s) and song) would go a long way. Nice alternative to my hype machine addiction.
It's a play button, and when you press it it actually just starts playing cool music? I didn't know that sort of straightforwardness and intuitiveness was even possible. Why can't the whole Internet be like this?
Very good execution and a good choice for the domain name. The design (green/white/red/hearts) reminds me of http://hypem.com/ - you should check it out if you haven't. It's got a similar theme (crawl indie blogs for music) and has a pretty big following.
The big problem with hypem I've found is that the popular tracks is completely under crowd control and wheres I used to use the service to find music that aligned well with my tastes (I discovered Phoenix and some incredible dance remixes of music I love through the service for example), now that the site's got more popular the music has following the American trend of Skrillex like 'dubstep' that not even the 'no remixes' setting can remove for me.
HypeM.com actually has a fixed set of blogs that it gets music from. I mean, I guess you could say it crawls those posts on those blogs, but it just doesnt crawl 'the web' per say. From my knowledge, HypeM has a relationship with the blogs it takes music from, in order to keep it's quality up.
I love it. I really wish there was a link to more information about the artist or at least to where the crawler found the audio. I'm sure the artists would appreciate it also. On that topic, how legal is this? Is all the music public domain?
Awesome site. Im curious as to how you defined your heuristics? Finding good obscure music seems like a hard thing to do programmatically.
It seems like you would want to find "good" music (which maybe could be determined by how many unique references link it) but thats the opposite of obscurity.. So how did you determine the quality of the "obscure" music you crawled? By hand?
I'm loving every single song it's selecting for me. Is there a way I can login and save my loved songs (apart from scrobbling) ?
So I liked a song, then I clicked scrobble , it redirected to last.fm - fine, but when it returned to mixest, it started playing another song, and also lost all my previously loved songs :(
Ah, that's because www.mixest.com and mixest.com are technically different domains :(. I'll get that fixed up, but your loved songs should appear on one of the two domains.
A small bug : sometimes when I click Next, nothing happens, so I clicked Next many times. That ended up in the song history becoming like this : http://i.imgur.com/yUcj5.png
im going to catch shit for choosing www over no-www, but if you want to make it so everybody is funneled to ONE domain (ie. www.mixest.com) put this in your htaccess and your problem will be solved.
I love this. There's a particular song that I absolutely can't get enough of, but it isn't labeled correctly (says "Track 4 by New Artist"). I've "loved" it but do you have any suggestions for finding out what it is?
This is awesome, I've been using it for the last couple of hours now really nicely done. Would love a blog post explaining the development process and the choices you made.
I have fond memories of discovering nice indie music with something called Indy, which was released in 2005 by Ian Clarke (of Freenet fame) <http://blog.locut.us/2005/04/17/indy/>; and featured a nice recommendation engine. Sadly, it seems to be dead now...
I definitely prefer it to Grooveshark radio. Nice and simple. The interface is a bit confusing, I pressed the music button thinking it was broken. Maybe add tooltips? Also, using the URL for playing certain songs is interesting. If a user bookmarks the page though, doesn't that mean they'll get the same song every time?
In both Firefox(ESR) and Chrome(v22) I am getting a message that I don't have the latest Flash installed. When I click on the message, adobe says I already have it. (no flash blockers enabled)
Okay, fought this for 10-15 minutes and discovered that it was my security settings in Firefox and my c-t-p settings in Chrome that were breaking things. Please switch to HTML5 or make the flash element visible.
And now that I have it up . . . how do I say this politely? This is the music of my nightmares. What genre is this? I know I've been stuck in a hole listening to classic jazz and hip-hop recordings for 10+ years, but this just sounds like dissonance and apathy. Most of the tracks I've found are less than 3 minutes long, about as tonal as Shostakovich and a bit croony?
Though, now that I typed all that, a single, solitary hip-hop track popped up after 20+ songs, and it's rather nice. So confused.
Works on iphone (without autoplay of course). Oddly the skip and play icons are reversed, but the functionality is not? Almost made me give up thinking iOS might not be supported.
i'm curious, how are you seeding your crawler and determining attributes for the next song recommendation? from my understanding, pandora has a team of musical experts and it takes them 30 minutes to gather data for each song (attributes/input to their recommendation algorithm).
i'm building http://cloudplay.fm a multi-source music player, and would love to have an API to Mixest. the world does need more niche'd Pandoras :)
This is truly outstanding. Please allow me to create an account so I can come back to my favorites (or make the UI clearer as to what is saved for me and what's not).
great execution for all the reasons already mentioned.
since it's layout is already simple and modular, a couple of meta tags and @media styles in your CSS would be good to make it collapse down on phones for easier access to the controls
I'd say with the quality of music I've seen, people would be willing to pay (month/year) for the service provided he develops it a bit further. Another possibility would be to have referrals to purchase the music like turntable.fm does.
You may have seen Mixest on HN 3 years ago, back when it was written in PHP. Its data was destroyed last year by a PHP virus, so I rebuilt it from scratch.
I've been using it as my hacking playlist, and I hope you guys will enjoy it too!