hello everyone, I am to the tedious task of extracting .wav\'s from the melodic percussion files and it is really starting to make me mad. I have wave surgeon and soundforge but I can not for the life of me get it to acurately extract the .wav\'s. It seems like no matter what I try it either divides to much or too little. Does someone out there have a sure fire way of doing this. I need specific\'s; like max and min threshold in wavesurgeon or the same in soundforge.
Having finished a 350 MB handbell set, I can appreciate this particular problem. I spent a lot of time before I could automate the edit process, enough that I question if doing it by hand wouldn\'t have been faster. Maybe the results will be helpful to you though. I\'ll put my soundforge settings for autoregion and autotrim at the end..
I went with soundforge after testing wavesurgeon and finding that the promising regioning function didn\'t seem well suited to my purpose (seperating a stream of percussive attack type samples). Wavesurgeon seemed to have that function for people to extract loops/breaks from songs. However, when I later asked Dave Govett from Nemesys about an inhouse program he used on gigapiano for sample seperation, he replied that he now uses commercial tools and suggested I try Wavesurgeon. It\'s possible I needed to spend more time with it, but with my limited eval of Wavesurgeon, the results weren\'t usable.
In soundforge, I found three functions applicable to sample splitting - Tools:Auto Region, Process:Auto Trim/Crop, and Tools:Batch Converter and Run Batch Script.
Auto Region along with Tools:Extract Regions should have been sufficient, but here are the problems I encountered. First, too may regions were being created(silent regions for example, as well as the desired single sample regions). I was specifying \'use release point for end of region\' along with the release sensitivity, so that the sample region would have its end where the sound level became very quiet. The side effect is that assumiung this is done perfectly, there is now also a silent region created from the end point to the beginning of the next sample region. I decided it was best to not check \'use release point for end of region\'. This way, sample end point would be at the same place as the begin point of the next sample. I decided I would it would be easier to deal with finding a correct sample end point later. Now the second problem with auto region is that I could never get the sample region to begin exactly where the sample truly began, no matter what I tried. The region always seemed to begin about 1 millisecond before the sample. I gave up on the auto-region as a one step tool and decided that the exact start point I would also find later. What the autoregion would accomplish for me was to split the long stream of samples into numerous single sample chunks, each of which would contain a very small silence before and a longer silence after the sample. If I then did \'extract regions\' I could turn the chunks into files and process the files further with the batch converter.
Here\'s where Auto Trim/Crop comes into play. Autotrim does a good job of both cutting out silence before a sample and after a sample. I found in fact that autotrim did cuts of silence before the sample exactly as I would by hand, when I AB\'ed a few samples done by hand and with autotrim. No more 1 millisecond of silence before the attack nonsense like autoregion produced. Cutting the silence after a sample was a little tougher to do right. When I recorded my samples, I used different mic preamp settings for the different velocities, thereby creating a high noise floor on low velocity samples. Assuming you did the same, you may want to save different settings of the release threshold autotrim for different velocities. I actually did the end point cuts by hand, listening for where I wanted them, but it\'s of course much slower that way.
Now given sensible settings of autotrim/crop which you have saved, you can automate all trimming using the batch converter. Add all the files you want to operate on(perhaps a velocity folder) and under Tools:batch converter ptions:conversion settings you can choose a class of \'Process\', function \'autotrim\', and an appropriate autotrim preset which you saved earlier. Then save this simple batch script, choose a destination folder for the processed files and run the script. You\'re done.
One seperate issue you may consider is normalizing. I decided against it but I know some people normalize all their samples. My feeling is that normalizing may make the samples less even, and since the process can be done by end users but not reversed, I left it alone. Normalizing uses the peak level of the sample to determine how much the entire sample level should be increased. But peak level is instantaneous so this is indicative but not reliable for the loudness of the sample. Obviously others have different opinions on normalizing, since I think many people do it.
Another issue- if you are pressed for space to include everything on a cd, you can use shorten (my favorite, since it self extracts, and has best compression and speed) or WinRAR with multimedia compression set, or wavezip(which I don\'t think selfextracts). These are all lossless compression. If you use shorten, to do gigs you have to choose 16 bit signed PCM, discard of 0, channels 2, and sometimes play with the alignment. I have found an alignment of 2 bytes allow it to work w/o error.
I hope this helps. Please if anyone has different methods or opinions for sample extraction, add a reply. This is the best way I know how to do it, but it seems like it shouldn\'t be nearly as complicated as this.
On another note, I have wished to distribute my handbell set to anyone who wanted it more or less free of charge(cd media cost and postage), but I can\'t deal with it for the next three months while in New Zealand. If you think you might be interested in a free extra with your set, let me know - hurchalla @ yahoo.com . It is 340 Megs and about 130 compressed.
soundforge sample settings I use as decribed above-
attack sensitiv 60%
release sensitiv .1%
min level 60%
min beat duration .1
check boxes are all unchecked.
function - remove edges outside selection
attack thresh -38 db
release thresh -48 db
fade in and fade out of 0.
Why couldn\'t you ahve posted this when I was editing over 3000 lines of Dialog (each to be edited in its own file AND each lie having up to 5 takes!!!) It took me forever to figure ou t a good batch process...and it wasn\'t nearly as good as your method! I jst wish autorename was a little easier to deal with...
If you need to extract individual wavs from an audio CD, try the following procedure:
1) In Wavelab: Open/Import audio track
2) In Soundforge, Auto trim/crop, attack thresh = -30 Db, release thresh = -40 Db, minimum interphrase silence = 0.1, fade in & out =0
3) In Soundforge, Auto region, attack sensitivity = 50%, release sensitivity = 50%, minimum level = 0.1%, minimum beat duration = 0.1, do not check use release point for end of region
4) In Soundforge, Extract region