I still have a lot to learn. Fortunately, I’m making mistakes with my own work, not with other people’s…
I tried to use Martin Zuther’s compressor, Squeezer, but you it crashed my Ardour. In my Antix 19 partition, where I’m still working there is the Ardour 5.12 version, the MX repositories haven’t been updated to version 6 yet. In my PCLinuxOS partition I do have the version 6, I tried Squeezer but it crashes too.
I used the LSP compressor, I hope the parameters are ok. I used a bus for it.
I think I like more the version where I use the compressor, but I think it affects a bit, although positively, the sonority, I don’t know what you think.
Lovely stuff, @Aleph! You are definitely in the right ball-park now with those loudness/peak numbers! What confuses me is that the numbers on both loudness analysis htmls are the same. If you’ve applied some parallel compression I would expect the numbers to change a little. Also, I note that most of the tracks are exactly at -20 LUFS which seems odd. Are you sure you imported your 32-bit float files to your “album session” without normalizing for loudness? And is your final export now trying to normalize loudness for the whole album or still per track? Finally, experiment, if you haven’t already, moving the compressor output between -15dB and -5dB and see what you like best.
Well, I did it again, but I got a similar result. Maybe something’s wrong, but I don’t know what it could be.
Here’s the analysis of the individual tracks converted to 32bits before importing them into the final session:
And here are the WAV files with the analyses.
With Bob Katz compression:
Now in the output of the compressor I only applied -7 db, I think it’s better, although at the moment I’m not sure either…
I add a screenshot of how I did the final session:
The loudness analysis you present shows 8 “ranges” which tells me that is not the analysis of the session WAV. The ranges appear for all the world to have been normalized to -20 LUFS or -1 dBTP (which ever comes first). That’s why 7 out of 8 hit -20 LUFS because their peaks are lower. The one track that doesn’t hit the LUFS value has high peaks so is normalized to -1 dBTP instead. You need to export/analyze only the “session” time span (and generate a cue file). Then, in fre:ac or similar, generate your FLACs, MP3s etc from that wav/cue combo.
If you want to generate all the individual files staying within Ardour, you really need to use one of the loudness meter plugins and add gain trim to your master bus to hit the desired LUFS or true peak for the whole album. Then export without any normalization!
The cue file generated only added the cuts at the beginning of each track, which was where I put a CD mark, being separated the tracks in the session you can hear a quite long silence at the end .
What should I do, remove the range marks, join all the tracks and keep only the CD marks to generate the CUE file?
You add CD markers and then when you export your “session” timespan just be sure to generate the CUE file. Download the audio converter “fre:ac” from here, drag in the cue file and generate your various file formats! I use Kid3-qt for my file-tagging in this instance.
Again, almost all of this can be achieved directly in Ardour if you use Robin Gareus’ EBU R128 meter (or similar) and gain trim the master instead of normalizing on export. I think the only thing you’d need to do post-export, if you so wish, is to add in album artwork in an external tagger.
You should be spacing your tracks as if you’d hear them on a CD, i.e. around 4 seconds, maybe, between movements, 12 seconds in-between whole pieces etc. In my experience that’s how the classical labels produce their FLAC files so that when imported into a gapless player everything flows as per the CD. Use the session “end” marker to set the final bit of silence (which on a car CD player would the be silence before it cycles back to track 1 )
Thank you very much, @bachstudies!! It would be very good if Ardour were to implant this possibility.
Having the possibility of doing crossfades in the style of Pyramix as you mentioned in the other thread would also be very good for us, the classics…
Well, I think everything’s almost done. (I hope…)
I’m uploading the tracks and analysis again. If you can hear it a little bit and give me your opinion about how you hear it I would also appreciate it
Individual tracks analysis:
dom jul 19 00:44:36 CEST 2020
File Int (LUFS) LRA (LU) TP (dB)
01 - Ripples Over Time - Michael Thomas - Luis Orden.wav -22.0 13.7 -8.0
02 - Monólogo - Miguel Ángel Gris - Luis Orden.wav -23.4 18.7 -6.4
03 - A lo lejos - Santiago D. Llopart - Luis Orden.wav -25.7 15.4 -10.1
04 - Trazos - Manuel Castillo - Luis Orden.wav -20.0 20.6 -4.7
05 - Capriccio - Roberto Gerhard - Luis Orden.wav -22.8 14.8 -1.0
06 - Soliloqui - Joaquim Homs - Luis Orden.wav -21.6 16.4 -7.0
07 - Preludio y Cante Op.34 - Preludio - Román Alís - Luis Orden.wav -24.5 16.6 -5.6
08 - Preludio y Cante Op.34 - Cante - Román Alís - Luis Orden.wav -23.5 13.9 -7.4 ALBUM-22.717.0-1.0 MMAX-7.8 SMAX-10.9
What I have recorded is very little for a physical cd, only about 35 minutes, I should record more pieces, although I’m thinking about uploading the CD only in digital to several platforms through CD Baby. In March I am going to record a new CD with my guitarist for the Spanish label IBS, they are doing very good works, it is the same label that cellist Asier Polo has recorded with, I talked about it in the other thread. Having two physical CDs in the same year I don’t think I’m too interested.
This is an interesting topic for me these days. The two DAWs that make classical editing fast (and enjoyable) are Sequoia and Pyramix. These provide not just source-destination editing but also a fully-fledged crossfade editor. Tutorials are few and far between and don’t seem to do justice to not only how much time is saved but also how clean the edits can be.
This all said, IMHO, Ardour does a better out-of-the-box edit than any other “standard” DAW because of the default short fade and the transparency on overlap. It takes way more time but the edits can be very very good. As I have been discussing with Robin Gareus, also very important to me is loudness conforming at the editing/mastering stage (coming soon to Ardour ) and the multi-format export which is absolutely stunning in Ardour (Pyramix has the amazing “album publisher” and recently Sequoia added this in a kind of half-baked way).
I keep a copy of Pyramix around just in case one day I’m hired as an classical editor (they all seem to use Pyramix or Sequoia) but otherwise, I’m a very happy camper on Ardour in Linux.
Hmm, a tricky one. My own feeling is that pushing past -18 LUFS integrated just to be competitive is a bit of a problem for classical. I imagine on various of your tracks you’d be needing to add far more compression than you are comfortable with for solo flute. I much prefer sticking to -20 LUFS (and ideally everything should eventually move to -23 LUFS). Recently I’ve had great success using a K-meter approach but with LUFS and set the momentary max to around -16 LUFS after analyzing the entire album. It works out to the -20 LUFS integrated ballpark most of the time with -6 dB peaks for quieter instruments (which seems to be another recommended practice).
If I was mastering pop, it would be different, I suppose. Even then, I’d probably just stick to Ian Shepherd’s advice to mix to -9 LUFS short-term max and call it good. Our classical music isn’t in danger of being turned down and, frankly, people understand that respected classical labels have never really gotten into the so-called loudness wars. If anything, classical seems to be a perfect match for -23 LUFS: