I did some tests with some songs and VoiceTracks, and used Cads IVP 5.1 with two global settings changed (for me, DUCK_INTRO_OFFSET=0.3, and DUCK_FADE_DURATION=0.400).
mAirlist v3.1.4 build 854
I found that generally the consistency was better this time and exporting to WAV produced better results. In a 5’10 playlist, MP3 320k came out 0.5 secs shorter and some datum points checks showed a 100-200mS variations between WAV and MP3. (Song 1 - Jingle - VT - Song 2) In reality acceptable for what I do.
Cad’s IVP5.1 worked well. (except some workflow issues that I always forget to do…)
The current mixdown algorithm is basically container playback into a file. However, part of the segue handling depends on messages that the audio source objects send to each other. That works very well for playback, but there can be delays when the processing is faster than realtime, as with mixdown.
Perhaps it’s possible to slow down the mixdown (using some parameter) and get better accuracy.
As mentioned earlier, the whole mixdown code will be rewritten for a future version, but that hasn’t happened so far.
[quote=“Torben, post:2, topic:7061”]Thanks for the feedback.
The current mixdown algorithm is basically container playback into a file. However, part of the segue handling depends on messages that the audio source objects send to each other. That works very well for playback, but there can be delays when the processing is faster than realtime, as with mixdown.
Perhaps it’s possible to slow down the mixdown (using some parameter) and get better accuracy.
As mentioned earlier, the whole mixdown code will be rewritten for a future version, but that hasn’t happened so far.[/quote]
That makes sense, I usually found that the 320k MP3 was closer to what I was after. It creates slower than WAV. That flies
One thing I did find, the accuracy is worse on short files such as jingles. I had a 6 second jingle, fairly tight when played live to next track, but on WAV export, seemed to start the next track 300mS later, and on the 320k MP3 export, ~500mS later. Redo the mixdown and the gap varies but within 20% or so.
(except some workflow issues that I always forget to do…)