What is the optimum bit rate and format for a VO demo file?

[Nikita]

Hi folks!

I've just joined your community and decided to start with a question that some of you may find too primitive. It's also quite possible it has been discussed here before. If so please forgive me, give me a link to the discussion and feel free to delete this posting. I've tried to use Search for 'bit rate' but all the numerous results I got were about money matters... ;)

So my question is:

What is the optimum bit rate and format for a VO demo file? I would like to have it as small as possible but at the same time to retain enough sound quality to show the product under the best angle.

I'm thinking about mp3 with variable bitrate and the following parameters:

Channels: mono
VBR Quality: VO («245 kbps)
Minimal bitrate: 64
Maximal bitrate: 128
Sample Rate: 44100
Encoding quality: 2 - High quality

What do you think? Is it possible to use maximal bitrate lower than 128?

[Mike Cooper]

Hi Nikita!

Not too primitive at all; that's what we're here for…

Personally, if it's a dry voice track, I always use 128Kbps constant bitrate mono. To me this is the right mix of it being good enough to hear some of the nuances of your voice and the ambience of the room, but not good enough to use professionally (and therefore unlikely to get me ripped off).

To my ears, 96K or lower start to sound "swooshy" and squashed, with that "underwater" quality.

The important thing, I feel, is that the quality threshold isn't so low that the artefacts in the recording get in the way of someone listening to your voice. Once they become distracted, you've lost them.

Voice123 and some of the other sites ask for Mono, 128K, if that helps.

Mike

[Paul Plack]

Welcome, Nikita!

I would echo Mike's comments, differing only on his observation about quality standards. Here in the states, audio for radio/TV, and certainly anything for websites or telephones, would be considered good enough (sadly) to use at 128K. If you're concerned about protecting your work, use a watermarking technique of some sort.

Anyone who routinely hires professional voice talent via the internet is accustomed to receiving 128K, constant bitrate files.

Are you bandwidth-limited on your end?

[Mike Cooper]

Here in the states, audio for radio/TV, and certainly anything for websites or telephones, would be considered good enough (sadly) to use at 128K.

Gulp. :sick:

[Paul Plack]

Gulp. :sick:

Mike,

I don't know about BBC, but audio processing in radio here is horrendous. Many of the big group-owned AM stations are rolling off audio at 5 kHz intentionally, to reduce nighttime interference, and the multiband compression is so overdone it makes everyone sound like they have sinus infections. Under such circumstances, 5:1 digital compression isn't even the weak link.

We have satellite radio here with music formats being compressed to 48K in stereo, and passed off as "CD quality."

Gulp, indeed.

[Nikita]

Hi guys!

Thanks for the welcome and your quick responses.


Personally, if it's a dry voice track, I always use 128Kbps constant bitrate mono.

Mike, and what bit rate would you recommend if it's not just dry voice and the track also includes some sound effects or music?




To me this is the right mix of it being good enough to hear some of the nuances of your voice and the ambience of the room, but not good enough to use professionally (and therefore unlikely to get me ripped off).

How can I be ripped off if I use for demos my previous projects done for various customers in the past? I suppose they can't be used for anything else by anybody else. Do I miss something?




Voice123 and some of the other sites ask for Mono, 128K, if that helps.

I've noticed that whatever you upload to Voice123 which is higher than 96 is converted to 96. It seems they prefer this bit rate.




If you're concerned about protecting your work, use a watermarking technique of some sort.


Well Paul, when I send my demos I'm not concerned about protecting my work as I believe those demos can't be re-used in any way.



I'm concerned about protecting my work when I send a project to a client not being paid yet. Usually they ask me to send a recording in WAV format. In order to avoid being ripped off I usually use the following method:
I make a copy of the WAV file and convert it into 96 (or lower) mp3.
I also cut off a small part of the track.
The resulting file I send to my customer just to give him a general feeling of the work done.
After receiving of the payment I send the original WAV file.
This is the method I worked out myself (© NK Translations & Voice-overs (http://www.translatorscafe.com/cafe/profile/VoiceOverDscr.asp?LinguistID=3063) ;)) .

For the moment I know nothing about watermarking technique used for protection of audio files. Could you please direct me where I can learn the basics of that?



Are you bandwidth-limited on your end?

No, I'm not: I've got a 24/7 broadband cable internet connection but sometimes I have to send 5-10 demos by e-mail and thus I want them to be as small as possible.

Guys, I've noticed that you both prefer constant bit rate. Why so? Why not to use VBR? I think using VBR gives you optimum quality/size ratio of mp3 files, does it not?

[Jacoby]

My first demo on my website five years back was actually a 96 kb/s stereo mp3... and it sounded fair. These days, I wouldn't go lower than 192. And given that a potential seeker probably has a MUCH faster connection than he did five years ago, a 192 is downloaded in a few seconds. But with a good mp3-encoder, 128 can sound pretty decent, too.

My best advice is to experiment.

[Mike Cooper]

I agree with Jacob - if there's music (and moonlight, and love and romance…) then I'd go for 192.

Voice123, I've noticed, asks you to upload at 128 but then converts to 96. Never been sure why…

As for getting ripped off, I thought you were talking about new demos for new jobs, rather than an existing track.

Watermarking? A search in most of these forums will bring up a wealth of information. In a nutshell, you add something (blips, intermittent background noise, etc) that renders the file unusable. You can't be ripped off, but many voice seekers are insulted by the idea that you'd think they might and may dismiss your demo without even listening to it all the way through. It's controversial.

One observation: if your customers are asking you to send a WAV, and you're sending an MP3 converted back to a WAV, they may think that's the best quality you can muster. It's unlikely that they can't play MP3s, so if they've asked for a WAV there's probably a good reason for that.

[jsgilbert]

I think the confusion may be stemming from the use of the word "demo" perhaps to innacurately describe an audition. This practice seemsto have been created by one or more ofthe onloine casting services.

Generally the terms are

Demo: One or more examples of a talent's work, often within a specific gnere and often used to show range, skills, acting ability. Demos are most often stereo and will (depending upon the genre) contain music and/or sound effects and mya even have the voices of other actors present

Audition: Most often, is a mono voice only recording of the talent reading a very specific piece of copy; either a part or total of the actual script that the client wants performed or something very similar. Unless specified, this would not include music or effects.

As for watermarking, this is not generally practiced by any of the talent agents or casting directors I am familiar with and seems to be exclusive to the online casting services, whereby the client and client credentials may be in question and/or the name of the client is kept hidden from the talent. Many talent believe that their auditions will be stolen and used without them receiving compensation or without their knowledge. While this doestake place, it also takes place via casting directors and talent agents as well. Many times the things that talent read for are not fully approved projects and often the voice is used for "pitch" purposes. Online casting may factor in students and amateurs looking for a "freebie".

The practice of watermarking one's auditions may prevent the ocassional audition from being poached, but also sends a very "wrong" signal to the bulk of the hiring community. I can safely say that none of the work I have been involved with would have cosidered any auditions that had been watermarked.

Generally, professional actors will not watermark auditions.

Demos are most often at 256k stereo 44.1 wav 16bit, aif or MP3. MP3 is a "lossy" compression technology and usually if it is used, an asrchival version in wav or aif format is retained. Often the "masters" or actual recording is done at 24 bit.

audiitons are mostly sent as 128k, mono, 441.1 16 bit MP3 files. Ocassionally as 96k, but lower than that is rarely employed.

[Scott Pollak]

Either my ears are plugged with horse manure, or you guys are too audio-elitist for the real world.

128k mono sounds awfully good.
256k stereo does, too.

How often is the voiceover you provide to the client ever used by itself, with no music or SFX, and listened to in an utterly quiet, sound-perfect listening environment? I'd say.... never.

Your voice ends up on commercials or e-learning or audiobooks or training videos that are dubbed and sent out, complete with music and full soundtrack, then listened to over FM airwaves in moving vehicles with road noise and honking horns, or on tvs blaring in acoustically crappy living rooms over barking dogs and screaming kids, or via some tinny 3" speaker on a 21" tv in a conference room that has industrial HVAC running or in a classroom with kids chatting, teachers teaching and traffic whizzing by outside the windows.

128k sounds just fine.

[Jacoby]

128k sounds just fine.

... right up until it get's severly compressed by the producer of the commercial it was intended for, and again processed by the end user, either a TV- or radiostation. It may even have been processed by the voiceartist before it went to the client - some people do that.

Here's how this could sound in real life. The sample is the same read, first part originaly saved as a 128 kb/s mp3, second part originaly saved as a 256 kb/s mp3. Then a typical production processing is done, and finally another portion of typical Orban or Omnia level of processing is done.

I clearly hear a difference. And may I remind you, I've used my own presets for processing and know them by heart. I didn't overdo anything to make a point. This is a feasible example.

To avoid any further & degrading mp3-encoding, and because the forum don't allow uploading wav-files, please download from here:

http://www.provoa.com/mp3test_128vs256.wav

[Colin Campbell]

I use 128k constant bit rate for mono voice, 192k constant bit rate for produced pieces.

Scott... 128k stereo does produce audible artifacts to my ears. Wether or not it will be presented in a noisy environment, I just can't let it go that way.

Of course, upon request I provide 44.1/16 bit (or higher) ucompressed PCM WAV or AIFF files and post them behind my web site for download to the customer.

If someone is going to steal an audition and use it without paying for it, I don't think the 128k is going to bother them one bit. I'm sure quality is not paramount to such an individual.

[Lee Gordon]

I believe both Voices.com and Voice123 re-encode the demos you upload to them and 96k before making them available to job posters, so if somebody is going to steal a demo from one of those sites, it wll be even lower than 128k but your point remains the same. The low bit rate is not going to bother them. Nor, apparently, will their conscience.

[Paul Plack]

How can I be ripped off if I use for demos my previous projects done for various customers in the past?...I have to send 5-10 demos by e-mail and thus I want them to be as small as possible...Why not to use VBR?

Nikita,

Sorry for the misunderstanding. If you're using generic demos created from past projects, no problem. But many jobs will want a custom audition based on a portion of their own script. This is where some people worry about rip-off.

If you want to send demos by e-mail, consider putting the audio file itself on a website, then provide the URLs of the audio files in the e-mail. That makes a very small e-mail that is still convenient for the recipient, and won't be filtered due to its size.

VBR used to cause compatibility issues depending on the MP3 codec used to create the file. Sometimes the files wouldn't play completely through, other times they'd exhibit odd noises or other artifacts. I haven't heard of this being a problem in years, but then, I don't know anyone who sends demos using VBR. I prefer to minimize the risk.

[Mike Cooper]

Your voice ends up … in moving vehicles with road noise and honking horns, or on tvs blaring in acoustically crappy living rooms … [etc.]

That's part of the argument to get the basic recording as good as possible, right? If it's crystal clear in the production it stands a chance of still being intelligible after all that lousy processing.

Whereas if it's crap in the first place and then goes through all of that, then its chances are roughly equivalent to a snowball's in hell.

[Scott Pollak]

I doubt anyone will tell you that anything I send them is crap, and I know you're not implying that, Mike. As a matter of fact, today I got yet another e-mail from another v/o talent asking if all the demos on my site were done in my home studio, as he felt the quality was (in his words) 'amazing'. And I'm curious: are you implying that 128k mono is 'crap'?

I've always felt that anything worth doing is worth doing to the best of one's ability, and of course that includes sending the highest possible quality of audio, both in terms of your talents and from a technical perspective.

However, it is also possible to carry such things over beyond real-world applications to an area where, in all honesty, a 128k, 44.1 mono narrative IS, in fact, plenty good enough for the real world. If you can send a 192k or higher mono file, go for it, but in the end I just don't see that anyone really listening is going to hear the difference. Most people are NOT listening to your end product in an acoustically perfect room with studio monitors and headphones and a frequency analyzer. They're hearing it in the environs I described above.

And not to downplay what we do, or the importance of doing it as well as possible, but perhaps an analogy would be the difference between, say, a Honda Accord and a Mercedes SL500. Both are nice cars. The Accord is no slouch, dependable, well-made, reliable, comfy. Obviously the Mercedes is, too. But if you basically need to get from point A to point B, both will do the job well. Only one will cost you 3 times as much, consume more gas, cost way more in insurance, and have a bunch of frills that -while nice to have - aren't truly necessary to accomplish the basic mission.

No doubt many here will choose to disagree with me.

[Mike Cooper]

I'm curious: are you implying that 128k mono is 'crap'?

Perhaps 'crap' is too harsh, and you're right, Scott: I'm not implying that anything you'd send out is crap either. Apologies if I gave that impression.

Listen, I've not got bat-ears by any stretch. I can tell the difference between most pop music at 128KHz MP3 and at 192KHz MP3, but above 192 I'm shaky. I'm really hard-pushed to tell the difference between 320KHz and WAV (OK, you got me; I can't…)

But I can hear the difference between my voice recorded as a WAV, and that WAV re-encoded to a 128KHz MP3 file. I hear it all the time when I submit auditions. If that MP3 file then goes through a chunk of comprod compression when an ad's being put together, followed by an aggressive Optimod on the way to the transmitter, or - worse still - goes through an Optimod and then gets re-encoded into a 128KHz MP2 stream like we use for most commercial DAB radio here in the UK, odds-on that it's going to sound worse than if the voice track had started off in full quality.

These things are subjective, though. Some producers aren't so bothered about the quality, and some are, even if the application doesn't demand it (like one of my IVR clients, who insists on 320Khz MP3 mono voice files for phone systems…)

[Scott Pollak]

Ironically, the studios at the radio station where I do p/t work are so friggin' noisy, especially with computer fans flailing away, that to my bad ears the voice work done there is awful. It's monstrously compressed and when I'm listening to a dry spot thru the headphones, you can hear the limiter shutting down between every sentence. While the talent is talking you can hear all the noise in the background, then for that split second between sentences all you hear is deathly silence.

Honestly, anything they get from me or any other talent HAS to be an improvement!

[Paul Plack]

It's been years since most radio stations had sound quality as a priority. Especially since they started buying up "market clusters" of stations, and trying to put six studios in one office space, managers who came up through sales and accounting looked at suggested floorplans, and couldn't understand why a studio had to be bigger than a phone booth.

Especially if it left less room for their corner offices.

A 128K mono MP3 is, if anything, is overkill in this environment.

My first job in radio was at one of the old Beautiful Music stations in the '70s. Great room acoustics, minimal processing, music library on 15 IPS analog reels (stored tails-out, of course,) and an RCA 77DX ribbon mic. Radio listeners under 35 have no idea how amazing analog FM, done right, could sound.

[Scott Pollak]

My first job in radio was at one of the old Beautiful Music stations in the '70s. Great room acoustics, minimal processing, music library on 15 IPS analog reels (stored tails-out, of course,) and an RCA 77DX ribbon mic. Radio listeners under 35 have no idea how amazing analog FM, done right, could sound.

OMG! That was MY first radio job, too! I'm serious. 1977, New Orleans, WEZB-FM, Eeeee-zeeee Radio. I had the midnight to 6 a.m. shift playing all the great songs by Henry Mancini, Frank Chacksfield, Peter Nero and so on. What a riot. Yep, had the big 15" tails-out reels, too.

[Paul Plack]

[quote=Scott Pollak;19942]OMG! That was MY first radio job, too! I'm serious. 1977, New Orleans, WEZB-FM...[quote]

"Every great city has one great radio station that plays Beautiful Music. Buffalo has WBNY! FM 96."

The processing consisted of a CBS Volumax limiter used without its companion Audimax compressor, and if we saw the Gain Reduction meter flicker more than a couple times a shift, it was considered an urgent problem. You had to turn up the car radio a little when you pushed that button, but man, those stations sounded pristine when done right!

"Xylene for the tape heads at every change, isopropyl for the rollers, and whatever you do, don't mix them up!"

Jim Schulke would be appalled at the thought of MP3 on radio. Even if the commercials were played at -3 dB from the music!

[Mike Cooper]

Indulge me…*I'm old enough to remember this, but it's been a while and my memory is playing tricks.

Tails-out is better because…? I have a feeling it's something to do with print-through, but I could be wrong…

[Lee Gordon]

You are correct. When a tape is played at normal speed it is more tightly and evenly wound on the reel and less suceptible to print-through (the magnetic image on a section of tape imprinting itself on the bit of tape flanking it on the reel, causing an unwanted, faint echo-like effect) than a tape that has been shuttled in fast-forward or fast-rewind and then stored. Therefore, the preferred practice is to play a tape from start to finish at normal speed, and then just leave it on the take-up reel without rewinding it. Then, the next time it is to be played, you rack up the tails-out reel on the take-up side, rewind it onto an empty reel on the feed side, and immediately play it. Once the reel has finished playing, it is again tightly wound tails-out and ready for storeage until needed again.

[Paul Plack]

Both forms of storage can allow print-through, but the tape manufacturers found that tails-out storage minimized what's called "pre-print," or a print-through signal which precedes the original. In other words, if a musical note prints through to the next layer out, you'll hear the print before the original, when it could be easily heard. "Post-print," or print-through which plays following the note, is less objectionable because it will often be masked by the decay of the original signal.

Also, since print signals take several minutes to form, and fade quickly when the layers are separated, rewinding just before use reduces the strength of print-through in either direction. Tails-out forces you to do that.

Led Zepelin's "Whole Lotta Love" actually used an effect which simulated both pre- and post-print. Listen to the effect on the portion of the lyric, "Way down inside..."

You can hear how the pre-print effect is much more distracting than the post-print, which just sounds like a delayed echo.

I know...:offtopic:Nikita, I'm really sorry we busted your thread so badly!

[Mike Cooper]

Pre-print! That's it. Thanks guys.

And sorry Nikita. Should I go and sit in the Naughty Corner? :'(

[Scott Pollak]

You guys are ignoring the most important part of storing the reels tails-out: It looks a lot better than when they're rewound and all unevenly stacked! :cheesy:

[Colin Campbell]

There's some debate as to whether this was "print through" or a deliberate artistic effect. Frankly, you'll notice that the "print through" does a different dance in the stereo field so I've always felt it was intentional. And, why is there no print through on the final... "loooooooooovee...?"

[Paul Plack]

Colin, I'd always assumed it was an intentional effect inspired by the sound of actual print-through.

[Colin Campbell]

Yes, Paul... went back and re-read your post. I see that now. But, it's interesting to play the clip for the folks.

[connieterwilliger]

My Fav band of all time...Zep...

[Greg Houser]

Here's a dumb question.

are you guys just interested in the MP3 playback bitrate, or are you concerned about recording something that will eventually be converted to MP3?

Depending on the response, we may have to chat about the PCM, sampling rate, and a nice fireside chat about the Nyquist-Shannon theorem.

I'm usually more interested in PCM and sampling rate, since it plays a greater role in playback (ragardless of format).

[Mike Cooper]

My vote's for "tape echo" - the stereo positioning of it seems to move around, too.

Er, we're still hijacking Nikita's thread… :offtopic:

Well done for getting things back on track, Greg!

One thing I heard was that you were better using 44.1 as a standard sample rate for anything that might end up being played back in Flash, as some versions of the Flash player get confused with MP3s encoded at higher bitrates. I've had clips on my site that have behaved "unpredictably" with clips that started off at 48K sample rate, regardless of the MP3 bitrate.

[Scott Pollak]

And as for getting back on track, at least for me personally, here's the order of preference I'm given by clients as to how they want the final audio:

1) .mp3 - 44.1k/128mbs (by a HUGE margin... probably 70% of my clients)
2) .wav - probably by about 20% of clients
3) .aif - for those on Macs
4) variation of .mp3, often as a 48k/192mbs file

[Jacoby]

Scott.... please tell me what mp3-encoder you are using... never heard of such an encoder, that does 128 MB/s mp3-bitrate :D :cool: :wink2:

Anyway..........! I'll play:

1) mp3 - 16-bit 44.1/320 kb/s CBR - this is what I give 80% of my clients, usually small-to-medium market radiostation, who probably couldn't tell the difference between 320 & 128 anyway. But often their "producer" is some young, hotshot DJ with a love for compressor plugins, and I don't want him to mess up my work completely.

2) wav - 16-bit 44.1 - this is what literally all my SERIOUS clients ask for: stuff to be used on big market radios, tv-commercials, DVD's for e-learning etc.

I've NEVER been asked to deliver anything else, neither different mp3-bitrates or aiff.

[JoeActor]

Ok, here's one for y'all:

I normally use the same settings as Jacob (above), but when I record music with multi-track sync needed, I use 48k sampling.

Why? Well, apparently there are a lot of sound cards that don't sample accurately at 44.1, but are fine with 48. Mine's one of em. I would lose about 1 second for each 30 of recording. Weird, but true!

Joe J Thomas
www.JoeActor.com (http://www.JoeActor.com)
www.SoundsGoodToYou.com (http://www.SoundsGoodToYou.com)

[Lynn Benson]

Hey, back to tails out for a sec...there was always a guy that didn't do it and the tape got loaded and played in REV.

(no, not me)

[Paul Plack]

There was a great story, may have been urban legend, years ago about a guy who made that mistake at The Voice of America. According to the legend, it was a one-hour program for broadcast on shortwave in Swahili, and nobody caught the mistake until days after it aired, when the mail started arriving...