Dealing With Too Many Samples

[Dr. Varney]

How would you go about constructing an audio book (play), in a DAW, which has many characters and scenes?

Okay - the history: I have been working on the first play for about a year and a half. I recorded nearly all the speech samples, editing them as I went along, for volume, pacing etc. I've ended up with about five different versions of just three scenes(there are 19 scenes in total and about 10 different characters) and about 400 individual samples, spread over about 20 folders, of which only a fraction are now in their proper scene folders.

Then I suddenly discovered I could slice samples in the playlist of the DAW and deal with volume and EQ, etc in one go, after the samples were assembled. This means I didn't need to do half of the work I've done, over the last two months - meticulously editing each samp......

You know what, I can't even explain any more of this. Basically, I've made myself an overcomplicated mess of folders, samples and back-ups all over the place and my brain is nearly fried. Basically, I have misused a comprehensive DAW as if it were Audacity Lite. Basically, it's going to take me the next 400 (it seems) years to finish at the rate I am working. It will be better if I just ask for your input...

From absolute scratch - start to finish, an hour-long play with ten different characters, all recorded seperately as rushes in .wav format...

How would you do it? Roughly, What would your workflow look like?

Bog-eyed and twitching with caffeine overload
Dr. V

 

[noisewreck]

I wouldn't record them all separately. I'd probably stick a mic in front of them, maybe put them in different rooms if bleed was an issue, and record the whole mess in one go. Then punch out/punch in issues.

Having said that, obviously now we're stuck with separately recorded takes, that are somehow delivered one by one.

First, I'd definitely come up with a consistent file-naming strategy, that would include the actor's name, character's name, date/time/take, scene and bit depth/sample-rate info (if necessary).

Then, in the DAW, I'd create a track for each character (for starters) and line up all the audio files where they belong on those tracks. Consistent file names would be a major help here.

Once the basics were there, I'd go from there as necessary, edit out/correct issues, and work on creating the scenes (via panning, reverb, FX, etc as needed).

 

[Dr. Varney]

Ah, I did start out recording them all seperately, then changed halfway and started doing long rushes, which correspond to whole scenes. Then I chopped them up into individual samples, with a view to loading each on a seperate channel, a mixer-track for each character and tackle each scene. In retrospect, probably a mistake (?).

It isn't possible to mic people up in one go because about two thirds of the characters are done by myself and the other actors had to visit me on seperate occasions.

If it wasn't for the sheer, mammoth volume of the recorded material, I don't think there would have been the same issues. Stopping and starting over so many times has been about my worst enemy.

Tonight, I just discovered it's a lot easier and quicker to load up one sample, per actor, per scene and chop it up in the pattern sequencer with the slice tool. Now I feel a fool for wasting so much time... If only I had realised the potential of the sequencer to do what has taken weeks to do manually, in just a few clicks! I've now got to reassemble all the short samples I spent hours cutting up, back into long samples, for this to work. Thankfully I've just found a way to export a sequence to just a few .wavs, using split mixer tracks.

I'm a tad more organised than my first post suggests (now I read back) and I don't mind complexity but rather enjoy the challenge.... it's just that I've made some really stupid mistakes up until now.

Perhaps the one thing I do have in my favour is (eventually) developing a sensible file-naming/folder strategy, without which, this could all have been a lot worse.

So now I've made some firm decisions about how I'll work it from this point on. I've just about got things under control and am now grudgingly going back over what I've done, to convert it to the system I intend to use.

But that said, I'm still very interested to see how others would tackle it because, you never know, fresh ideas might save me some headaches later on.

The good news is, that although I have done some test sequences of scenes, I decided quite early on, to organise all of my source material, before trying to sequence the play.

Dr. V

 

[SouthSIDE Glen]

One of the major advantages to digital editors is their non-linear, non-destructive nature. To translate that into grammar school English, There's very little need to chop files up into smaller individual clips until you have a million different files. Keep your recorded files as intact as possible - thus keeping the number of files you need to manage to a minimum - and let the DAW editing do all the work. It sounds like you have kind of discovered that eureka moment after the fact, but it's worth repeating here for the record.

I agree with Noisewreck that saving the recordings with an organized file name structure is a great start, perhaps with an act_scene_character_take nomenclature. I might also (this is optional) break the play down into a separate DAW project per act.

It's hard for me to make a suggestion as to how to continue with what you have now, it all depends upon just how bad things are in different respects. If it's really bad, I might take a day just organizing your files; finding out just what you have, putting them in the right scene folders, deleting all old/unnecessary files, etc. If, OTOH, it's really, really bad, you might just need to get through this one fast and get it done, learning your lessons for next time.

G.

 

[frederic]

I never did an audio book however I did all of the audio for a 45 minute training video, with three speakers interacting with each other so maybe this is relevent. The video was basically a slide show containing screen prints of the customer's proprietary software in use, the usual corporate statistics, policy information, how they define and organize their customers (enterprise accounts, mid-markets, etc) and various other things.

For each slide, the audio part was comprised of three speakers interacting as if they were having a conversation - a primary speaker who had the answers, and secondary speakers who would ask questions or expand on things - much the way the common infomercial goes these days.

Anyway, we recorded the audio for each slide separately and in numerous takes. Since the written words the speakers were to rattle off never included personal names, I chose to name the "characters" so I could keep it straight in my head, instead of speaker1, speaker2, etc.

So my tracks were named like so:

slide4-bill-take4.wav
slide4-bill-take5.wav
slide4-mary-take1.wav

Since most OS (and in turn software) alphabetizes directory listings I put the slide number as the first part, the speaker as the second part, and the take as the third part. When listed, related tracks would be together - all the slide1 stuff was together, all the bill waves related to sldie1 were together, etc.

For you, instead of slides, chapters might be a good indicator??

One of speakers in this project was a pain in the ass as she couldn't do her lines without tremendous issue - either she broke out giggling or seemingly couldn't read her lines. She easily had 50x more takes than the other two speakers who would nail it within two takes almost every time.

The three speakers were not professional speakers - one was a senior finance manager with a very rich, deep voice - he was the voice who explained everything and answered the questions of the other two. The second voice was a salesperson, and his salesy-enthusiam got on my nerves very quickly as he sounded excited all the time, and the third voice was an admin who had a very pleasing voice, but she struggled on every slide to read her lines clearly without breaking into a silly, giddy laughter.

It was a very long week.

We recorded this using four empty offices with wires draped between them, through a Samson MPL2242 analog console feeding an Akai DR8, to a scsi hard drive.

The reason why they wanted to do it this way was to make it more "fun" to watch then the typical powerpoint slide presentation, and also have it on DVD so it can be played anywhere - conference rooms, hotel "theaters", home equipment as each new salesperson was given a copy when they were hired.

 

[Dr. Varney]

Yes, Glen... I think that's damn good adivce. Since I recognised the way in which I could (and should) be using the DAW, I've got a heck of a lot less files to deal with.

Yes, I've spent two days, in fact, just re-organising - and it's paid off.

Because the DAW will export multiple tracks, split per mixer insert, I was able to compound all of the voices for each chapter (scene) into one whole wav. So now I have a folder for each scene and inside those, I have, for instance:

Scene 1
01_Alyssia.wav
01_Bast.wav
01_Molly.wav
01_Narrator

Scene 2
02_Alyssia.wav
02_Bast.wav.
02_John.wav
02_Molly.wav
02_Narrator.wav

...and so on.

So then I only have to load each wav into the sequencing mix and slice it up in situ, to create the breaks. I can easily space and pace as I desire; add volume envelopes where necessary.

Because of the "up to 64 of these and 128 of those" stuff I hear about DAWs, I originally reckoned that it would be impossible to catch the whole play with one project file so, prior to starting, I marked the original word document into managable scenes.

I then discovered from reading the specs that FL Studio has no limits whatsoever on the amount of instrument/sampler channels one can set up and, due to some complications I encountered with the mixer when jumping between scenes (yes, I was trying to add FX and 'mixter/muckster' the sound too early and, despite constantly saving strip/track states, keeping track of send knob-tweaks (which you CAN'T save) I'd made from one project to another became a nightmare) I realised only one project (in theory) was needed.

^Good god, that's a mouthful... Sorry!

Now, depending on how the computer itself performs, under the weight of a sample-rich project, I may split the job into 'acts', which consist of several scenes each. Some scenes are quite short, so I'm sure this is going to be fine for the most part.

Frederic, thank you for sharing your experience too. I found it extremely helpful and so it was a pleasure reading your post.

So yeah, I've learned the hard way not to twiddle with the mixer, until the general structure is complete - it's just that I couldn't wait to hear what the voices would sound like, with reverb and compression, etc. Down boy!

I thought this would be a quicker job than it's turned out to be, but I've learned SUCH a lot from the mistakes. Now I've made a plan and got the fragmented material from the 7 scenes I'd already done, tidied up and aligned with the new plan, the future of this project doesn't seem so daunting.

Cheers

Dr. V

 

[frederic]

You should always record something like this dry, controls neutral.

If you must twiddle knobs (even virtual ones) some systems allow you to put your outboards on the monitor chain thus allowing you to record dry yet not listen to it dry.

Typicially with speaking voices once you set things a certain way, they stay that way for a given vocalist.

 

[Dr. Varney]

I record everything dry, with the exception of the guitar.

I wouldn't dream of recording the spoken word with any effects. The issue was more to do with me having broken my first attempt into one scene-per-project. It meant, simply, that if I wanted to make a change to a voice in one project/scene, it was a nightmare to try and match the mixer settings from one scene to another. Of course, I saved all the mixer presets but you can't store send & return levels so, everytime I loaded an effect, I'd either have to resort to guesswork or laborious note-taking, which just slowed me down to a crawl. Imagine doing 19 scenes then, at around #15, deciding you wanted a little more reverb or compression on just one voice... Absolute nightmare! That's why I've started fresh, stripping out all the effects and plan to mix the entire play as one project.

As for volume, I've recorded at roughly -14db (though I use my ears rather than meters) to allow some overhead before I start adding compression and levelling envelopes (which will come much later).

Regards

Dr. V

 

[frederic]

I feel you're pain, however rabid note-taking is often what we have to resort to even though we have access to cool technology. The key to having that data be useful is organization, much the way you organized your audio tracks.

What I did for the training video I spoke of earlier is draw in "Viso" one of the samson mixer's input channels - knobs, faders, etc, and paste that onto the left side of the 8x10" page. On the rest of the page were boxes for speaker (instrument), mic used, channel number, scene, take, and so on. Below that and also to the right was a bunch of lines for freehand notes. As things progressed, I'd pull out blank ones and fill them out accordingly, then shove them into a binder in order.

Annoying, but I can't remember what I had for breakfast so I don't even bother trying to remember the settings of "effects send 3" knob.

 

[Dr. Varney]

Now I have a completely new problem... After combining each wav into a compound wav for each character, per scene - I can't possibly sequence it! Cutting the long wavs into individual pieces in the sequencing project has brought with it an almost impossible difficulty.

Back to the beginning - only this time, the complexity is even worse because I have spent time editing and polishing up some of the samples. Just renumbering them all is going to take a week!

I'm on the verge of telling the magazine editor to forget it. I've driven myself so far into a corner, I simply cannot see my way forward anymore...

Now I'm not exaggerating - if I were to do this as a full time job, 40 hours a week, I could not get this done in less than six months and I have to do this in my spare time. When I add up the time I've spent already, that's about what it amounts to: impossible!

All it is, is a sequence of different wavs of short bits of speech. You'd think this would be easier than making a song but now I am feeling depressed at how long I have spent already, only to realise the whole thing is a fools errand and all my creativity over the last year and a half has gone to waste .

Help!?

Is help even possible? I just don't know what to do with it anymore.

Dr. V

 

[SouthSIDE Glen]

Now I have a completely new problem... After combining each wav into a compound wav for each character, per scene - I can't possibly sequence it! Cutting the long wavs into individual pieces in the sequencing project has brought with it an almost impossible difficulty.

I don't understand quite what you're saying here, Doc. What exactly do you mean by "sequence it" and why is it giving you problems? Also, it may help if you give us the make/model of the DAW software.

G.

 

[Dr. Varney]

FL Studio 8 Prod. Edition.

Sequencing = Dragging in samples from DAW browser into DAW playlist, in the correct order and deciding how much silence should exist between sentences.

----------------------------------------------------

Okay, sorry... What I mean to say is: because I can't save all my meticulously tweaked send-levels data in the project (can only store mixer inserts and send contents, but not send levels to each track) along with the project, I decided to sequence (meaning drag samples into the playlist in the order they must appear) the whole play in one project, rather than split into scenes. However, this would mean a huge number of instrument channels to house the massive number of wavs the project needs... So I spent time compiling scene-long samples for each character, out of the individual sentences, which I would then chop up and move around - something like a word-grid puzzle, until everything is in place. It means I only need one wav per character, per scene. Is this making sense? I hope so.

Trouble is, trying to solve that 'word puzzle' is proving just too hard and more time consuming than ever. It's confusing and has a low margin of error. The mistakes I keep making outweigh the clever idea (or so I thought) I'd come up with. In fact, it's the stupidest thing you ever saw.

So now I'm back to chopping the scene-long speeches back up into sentence-long samples; that is: cutting them as one person stops talking and the other starts. But because keeping track of what I've done, plus the changes I've made to some of the scene-long wavs, I don't know where I am anymore, in this haystack of (beautifully organised) wavs. Proving that if you're too organised, you can go just as insane as if you were not organised at all!

Okay, if I roll my sleeves up and knuckle down, I could probably re-chop and re-save all the wavs again from the start. Only... then I am back to the problem of having a project which has too many wavs. It takes ages to load up the material for just three or four scenes - and there are about 16 scenes yet to address!

Then, that leads me back to the problem of: if I am gonna address this by sequencing the play as scene-per-project, then I have to work out some way of getting the send levels the same from scene to scene/project-project. I can't just master the whole thing as a two channel import because each voice needs different treatment and varying levels of send, from scene to scene. For example, I have very detailed changes in the amount of reverb, when a character moves from the outside, into a room or is thinking to themselves, etc.

Okay, I could take notes of send levels (I have) but when I decide something major needs to change (a macro change) across all scenes, in one voice, I have to correct those notes, open up all the scenes and tweak something repeatedly I'd rather tweak once, across the entire play.

Before you say it - no, I'm not trying to master (or even attempt to mix) as I go along. Yes, I set up a test mixer, to determine how each voice would be treated, but now I am concentrating on sequencing the wavs. Ideallly, what I wanted to do, was mix ALL the material, in one long mixing session, to make sure everything sounds like it belongs. Then I was gonna take a simple, two channel wav of it all - and master that to get the final polish.

Sorry - I was hoping I wouldn't resort to bugging you with this problem. I quite understand if you're bored senseless after the first paragraph... I've kinda gone numb from the neck upwards.

Thanks for reading.

Dr. V

 

[Dr. Varney]

Ah... Just thought... I suppose I would be looking for some way of compressing my wavs without too much loss and checking the "Keep On Disk" option in the channel settings? I assume that would speed things up?

(I suppose you can tell, I am determined to finish this damn thing, come hell or high water!)

Dr. V

 

[noisewreck]

Time to upgrade your DAW to something more feature rich. I think some of this workflow nightmare comes from the legacy Floops mode of operation (the word "Loops" isn't there by accident you know ). Life would be much simpler with Cubase or Pro Tools or Logic or Sonar where you can save all the mixer settings and presets (even per channel basis, at least in Cubase) and call up as necessary. Plus, you get much better tools overall.

 

[frederic]

Anytime you are overwhelmed it's time to break things down into smaller pieces. This applies to your situation as well as say, painting an entire house or any other project.

Focus first on scene/chapter one. Import, tweak, adjust, monitor, then send out a final mix of that scene/chapter as a stereo wav.

Put that aside, then start on scene two.

When you are done, you will have one stereo wav for each scene/chapter, and you can link those in your DAW sequentially or use something fairly simple (and flexible) - audacity. Then your final output would be a stereo wave file.

There's no need to put every scene/chapter into the DAW at the same time, torturing yourself trying to keep it straight. Small bites gets you there just as fast as one big bite, without the stomach upset.

 

[Dr. Varney]

Time to upgrade your DAW to something more feature rich. I think some of this workflow nightmare comes from the legacy Floops mode of operation (the word "Loops" isn't there by accident you know ). Life would be much simpler with Cubase or Pro Tools or Logic or Sonar where you can save all the mixer settings and presets (even per channel basis, at least in Cubase) and call up as necessary. Plus, you get much better tools overall.

I can save all the mixer inserts, just not the sends. These are vital to me, as they are the knobs which determine what proportion of dry and wet signal reaches the output. See, here...

I've coloured the sends in orange, the character voice inserts light blue etc. In this shot, the strip called 'SAMPLER' is selected and you can see the insert FX slots for that strip, to the right hand side of the rig. From the 'file/export' menu I can choose to export my project as 'bones'. What this does, is saves all of my channels and all of my mixer strips and their volume levels, plus all effects and their presets in a folder, from which I can drag any element I choose, into any future project. It's pretty good, except... see the send knobs (below the track sliders on the orange send slots)? Can't save where they are at. No matter what you've done to them, they will always load up at zero. The only way to preserve them is to save the entire project (or make a little picture like I have). Basically, I cannot save track routing outside of the project file.

That's pretty stupid, huh?

Now - will Cubase do that for me? I have a lite copy of Cubase which shipped with my soundcard. I don't mind learning to use it, if it will help.

Dr. V

 

[Dr. Varney]

Anytime you are overwhelmed it's time to break things down into smaller pieces. This applies to your situation as well as say, painting an entire house or any other project.

Focus first on scene/chapter one. Import, tweak, adjust, monitor, then send out a final mix of that scene/chapter as a stereo wav.

Put that aside, then start on scene two.

When you are done, you will have one stereo wav for each scene/chapter, and you can link those in your DAW sequentially or use something fairly simple (and flexible) - audacity. Then your final output would be a stereo wave file.

There's no need to put every scene/chapter into the DAW at the same time, torturing yourself trying to keep it straight. Small bites gets you there just as fast as one big bite, without the stomach upset.

Thanks. I reckon that's good advice and it's pretty much where I'm heading this time around, except I may split the job into 'acts' which comprise a few scenes (I'm not sure yet). As for saving those sends though... not being able to do that is why I thought to do it in one project. This is a problem.

Dr. V