Subject:voice editing

Posted by: devinemke

Date:4/16/2005 6:44:45 PM

I use SF8 for editing voice overs for IVR phone systems. the scripts I record are usually several hundred individual voice prompts, each with a separate number. I typically record the voice talent reading an entire script as one big WAV file then later go back and edit each indivdual prompt as a separate WAV file (then later batch them all to VOX). this all works fine.

anyway, my clients have very specific specs on the how they want the pauses in the voice prompts to be: there needs to be exactly 20 frames of silence between each sentence, and exactly 10 frames of silence for pauses within an sentece (like following a comma). up to now I have been simply opening up the big unedited WAV file and using the "Auto Region" feature to make it easy to select to pauses between phrases. i then listen to each prompt, select a pause and use "Mute", and then have to calculate (in my head) the difference between how long the pause is and how long it needs to be cut by. for example, if i select a pause (that is in a sentence break) and the length of the pause is 1 second and 7 frames then i know i have to select 17 frames and then select "cut" (to cut it down to 20 frames). this can get rather tedious after a while, especially on very long scripts. is there a quicker way to do this? ideally, i would like a feature (bound to a keystroke) whereby i can select any part of a soundfile and SF would automatically "mute" it and then cut it by the difference between the selected length and a user definable length (like a 20 or 10 frames). i looked at "Auto Trim/Crop" but this does not seem to do what i want.

this is really not something that could be done in batch mode because again there are different pauses lengths that i need to use (20 and 10 frames) and the batch converter will obviously not know which are sentence breaks and which are just comma breaks, etc... I realize of course that this is a rather odd feature request that probably no one else has ever needed to do but i thought i would still ask.

thanks for reading

Subject:RE: voice editing

Reply by: Rednroll

Date:4/17/2005 5:48:57 PM

Sorry, I'm not quite following what you're asking for, but here's a suggestion anyway. You can use the audto Region to extract each sentence out to it's own individual sentence, or do it manually if that works better. Then goto the end of the file and insert some silence. Create 2 more regions, one of 20 frames of Silence and the other of 10 Frames of Silence. Now assemble all the regions in the playlist editor, followed by the proper silence region. You could also do this same type of thing using CDA and assemble it all together in CDA, where you can create split edits, trim the head and tail and then insert the two silence wave files.

Subject:RE: voice editing

Reply by: Andreas S.

Date:4/22/2005 10:24:28 PM

Hi Mike. I do almost EXACTLY the same kind of work, though project requirements vary from xxx millisecs of blank to 0 millisecs (front and back) on the file.

I find using batch processing comes in VERY handy. I actually use a limiter (either a limiter filter, or more often, the noise gate) to clip out any audio (and breaths) below a given lvl. Then I use AUTO REGION to mark the regions and cut and paste the individual clips into new files. Then I use batch processor to either add the silence (or not, since AUTO REGION clips within xx millisecs).

In any case, I've uninstalled SF8 in favour of SF7b (which is fine on my system), but I've heard rumblings that Auto Region may not be included on it (I never got far enough in a project to actually find out - just got too frustrated...am now waiting for the first patch to create a working version of the program).

Subject:RE: voice editing

Reply by: sean@oregonsound.com

Date:4/24/2005 10:53:59 PM

An interesting topic, as I do some work along these lines as well---though fortunately none of my clients' demands have been quite that specific. I haven't figured out an automatic way to do this, but it occurred to me while reading through this that the best place to make this option available might be in the Extract Regions options. Similar to the option of adding "handles' to OMF or archive files in various software, an option to add a specified amount of silence at the head and tail of each extracted file would be killer. That way we could do very tight and fast editing in the creation of the regions, requiring no region-by-region calculations for insertions or cuts.

Subject:RE: voice editing

Reply by: mpd

Date:4/25/2005 6:57:29 AM

FWIW, here is my general workflow for VO work.

We use SF for recording and editing the VO for our training products, and other multimedia products. Typically, each cell in a script results in a different file.

When we bring talent into the studio, I record everything and just drop markers for each cell and have the talent slate each cell. Since SF only has one type marker, I mark the script where I need to make edits.

A script starts out as one (large) file. As I edit, I create a region for each cell and make sure the region name matches the cell name in the script. I don't worry about timing, but I make sure to mute out the slates, producer comments, etc between cells.

When I am done with the editing, I end up with a single file with a region for each cell in the script. At this point, I have not applied any processes or effetcs. I then extract all of the regions to individual files.

I then open a file or two and work on a post-processing plugin chain. Typically, the chain is EQ -> Gate -> Compression, and have a preset for a starting point for our more regular talent. I check the chain against a few of the files, and then save it off as a preset.

I then create a batch process with the following steps:

1. RMS normalize
2. Plugin chain
3. RMS normalize
4. Insert silence at beg
5. Insert silence at end

I batch all of the files this way. I then run auto trim/crop on each file by hand (because of a bug in SF8) to set the handles to exactly what I need for this particular project. The insert silence / insert silence / auto trim is the best method I have found for setting handles to exact lengths.

I then go back and listen to each file and tweak the volume as needed.

Subject:RE: voice editing

Reply by: sean@oregonsound.com

Date:4/25/2005 7:32:50 AM

Ahh, in all the years I've been using SF I never noticed that "insert silence" was an option in the Batch Editor! Thanks for pointing that out.

Your method is very similar to mine, except that I tend to process the audio before extracting the regions so I can easily check for any inconsistencies that may have developed during a session.