Crossfading support (non-equal power) - backend work

Paul added a comment - 02/Mar/11 5:10 PM Should we replace the fade in/out values with just one crossfade?

Douglas Arellanes added a comment - 02/Mar/11 5:15 PM A single value, while simpler for beginning users, may not be "professional" enough for advanced users. This is definitely one to consult Robert Klajn on.

Paul added a comment - 16/May/11 8:14 PM From the forum: http://forum.sourcefabric.org/index.php?t=rview&goto=10909#msg_10909 Re the proposal, first I don't understand the difference between options 1 and 2, as they kinda sound the same to me. re the question, no, fading in really is not used. The problem is that the term crossfade when used in this context really is a misnomer, as no (or very little) actual fading is done. Overlapping would be a better term. Unfortunately, your proposal will not work well as it is too simplistic. "crossfading" is more about starting the next item at the right time than anything else. You can't just use a fixed time point for the segway. Some tracks fade out, some come to a dead stop, and many have seconds of silence at the end and even at the beginning. I've used both the SQR Crossfade plugin for Winamp ( http://www.winamp.com/plugin/sqrsoft-advanced-crossfading-ou tput/32368) and the crossfading settings in Stationplaylist Studio ( http://stationplaylist.com ). Both have a trigger level in DB that the first song must reach before the second song will start. Both have minimum and maximum lengths of time from the end of the first track that the second will start. And both have a minimum DB level for both the beginning and end of the track in order to kill gaps. I'm still getting to grips with the Stationplaylist settings, but I managed to get the SQR Crossfader to sound pretty good, even to the point where I can play IDs back-to-back and they won't step on each other. I've not sampled many of the open implementations of crossfading out there. Ices0 ( http://icecast.org/ices.php ) has an implementation of what not to do. It has a fixed crossfade duration and both fades out and fades in. To get anywhere near a decent result, I had to set crossfade duration to 1 second, and even that is not ideal. I can't immediately find a stream where I know this is enabled, but set this up with a playlist and some short station IDs and you'll soon see why it doesn't work well. In fact, I actually turned it off completely on some setups where I'd originally had it turned on. Pytone seems to have a similar implementation, judging by the config file. I've not tried it out though. MPd can apparently crossfade, but I've not looked to see how detailed or configurable it is. I think I recall XMMS (version 1) having a crossfade plugin, but again I've not tried it. It seems that a propper crossfader has been on the XMMS2 Google Summer of Code todo list for the last few years, including this year. One system which does have very nice crossfading and dynamic compression is the system I was looking to use Airtime instead of - PRS ( http://sourceforge.net/projects/prsradio/ ). I don't know how it does it though - it seems to be hard-coded because I don't know of any way to configure it. I'm reluctant to use this software because it's largely unmaintained and sometimes its stability is questionable. I've been making enhancements to the web UI, but I'm not a C programmer or have any great expertise in audio processing, so the playout side is fairly incomprehensible to me. Still, it may help you in some way. I'm about to try deploying it on a new site, I can post the link if I get it going so you can hear what it sounds like.

Daniel James added a comment - 21/Sep/11 10:19 AM What's being described here is known as an equal power crossfade, rather than a 'fixed time from end/start of file' crossfade, as currently implemented in Airtime. Some background info at: http://emusician.com/special_report/types_fades/

Daniel James added a comment - 13/Mar/12 3:56 PM Liquidsoap supports equal power crossfades: http://savonet.sourceforge.net/doc-1.0.0/smartcrossfade.html There would need to be some way to calculate the new length of the playlist once it had been shortened by crossfading, but before playout time. In the example script linked above, some crossfades are skipped, depending on the end and start levels of the two files.

Paul added a comment - 03/Jun/12 11:43 PM From the forum: http://forum.sourcefabric.org/discussion/comment/17404#Comment_17404 He would like non-equal power crossfading: "I'd like to add my support for this too - in fact I'd like to go two steps further: First, rather than crossfading (fade out Song A, fade in Song B), I'd like to be able to start Song B at full volume as Song A is fading out, at about 5 seconds before the end of Song A. I *don't* want to fade in Song B, because if Song B has a nice punchy start to it, the impact is lost if it's quiet to begin with. I think this is a really key feature - if you listen to commercial radio stations, particularly in the UK, then sound is constant; they do not wait for the song to completely finish before starting the next. Secondly, you could take this one step further and have multiple audio files playing at once: Say, for example, that a song C has a long intro, and you had a spoken Jingle. You could start song C, dip it and play the long spoken Jingle, fading song C back up at the start of the lyrics. Wouldn't that be a pretty cool feature?"

Daniel James added a comment - 28/Sep/12 6:16 PM A visual crossfading editor needs to handle the file before and the file after the current one, so that fades can be managed at both ends of a voice file (e.g. for voice tracking).

Paul Bonass added a comment - 08/Feb/13 1:01 AM - edited I couldn't be more behind this feature! For me it's the only big thing missing in terms of making it sound (to a listener) no different from a professional commercial station. At the moment the fades just don't work well for what I'm trying to achieve. Many current songs end cold on a word - taking 'Britney B*tch' as an example in 'Gimme More', the current fade system would fade the word 'B*tch'. In a nutshell, Systems such as RCS Mastercontrol and Myraid all have a feature where you set a 'NEXT MARKER' on each individual track. This 'NEXT MARKER' tells the system 'At this point, start playing the next song (B) at full volume and start fading me (A)' Having this option would be a massive leap in the software bringing it up level with so many other applications and providing much tighter slick playlists. Is this still on track for a future release? Is there any workarounds at present? The work you guys are doing is amazing! So thankful!