This is an open-source clone in C# influenced by the LivePlaylists project by Yaniv Lerman. This is not a replacement of the original work; it's a side project I opened source to let streamers continue enjoy the service and encourage other curious developers to observe the code and learn from it, and perhaps influence others to pull off their own variation of the project.
How it works? Strategies explained...
The program monitors the channel every fixed interval in seconds; it then saves a chunk of 128KB in size (00:08 seconds in length) of the stream and then uploads it to Shazam for recognition (this strategy is used in order to reduce API rate limits and cost for fallback). After a few tries, if not recognition was successful, we fallback to AudD.io API for recognition as a backup strategy. When a song has been recognized, we do a little math here. Each strategy returns a data structure that contains properties such as the track duration and the current offset of the track, so a clever algorithm I developed was to subtract the track's total length and current offset and let the thread sleep before next execution. After all strategies were executed and no recognition has been successful (due to a broadcast or w.e.), then the method returns without doing anything and waits for next execution. It's important to mention that each execution shifts to the next strategy while respecting that each strategy has its own RetryCount property that tells the scheduler how many times to retry the strategy before moving to the next one. When the scheduler reaches the last strategy or we have a successful recognition, we reset the index and revert back to the initial strategy.
In the end, I decided not to maintain my own playlists. At least for now, because I do not believe it is needed as Mr. Lerman himself is hosting his own playlists with the PLAYingLIST.net project, and because of redundant financial it will cost me. I prefer to put my effort into maintaining the open source project instead of hosting my own playlists and servers.
Extending support to more channels is quite easy. You create a new class with the channel name (and "Channel" suffix) in the "Channels" folder and inherit from Channel. Then, you supply three properties. A url of the stream (StreamUrl), channel name (ChannelName) and the id of the spotify playlist (PlaylistId) initialized at ctor.
-
I was thinking of going hard and developing a special algorithm that will analyze the sound waves of the audio binary file in order to detect certain patterns and extract interesting parameters, like if the current stream is a broadcast, breaking news, traffic highlights, or a song. This is a clever strategy that could eliminate the need for a redundant API call, which could be cost-friendly. I have been reading a little here and there, doing some small research, and I think FFmpeg could be of use here. This is something that I'm still not certain about, but it's just an idea.
-
Implement Odesli; this provides direct link support to all major streaming providers. This eliminates the need to perform a redundant search and directly paste the song to the playlist without heading.
AlekseyMartynov - For the Shazam client shazam-for-real, and the inspiration from the profile picture.