Preliminary voice samples list.

[Destructavator]

The air... does things.

dispersion perhaps. different frequencies have slightly different speeds.

like thunder. close thunder cracks. distant thunder rolls.

It's called proximity effect - the same kind of issue to deal with when one needs to mic a guitar amp.  Up close it has more bass, too close it is harsh and boomy with SPL overload.  To explain any further would mean getting very technical.

Most professional studio vocal mic manufacturers (that I've seen, I've heard this can vary) recommend a distance of 1 to 3 (or 5, can't remember the exact amount) inches from the mouth to the mic, and also to position the mic so it directly faces the sound source (your mouth) and not held straight up.  All these things vary, of course with the type and class of microphone you use and what the mic is designed for.  Generally it is not recommended to "swallow" the mic (having the device close enough to touch your lips) or to point it away from your mouth at an off-angle.

There are also "shotgun" mics and microphones that are made for picking up far-away sounds, and other types, as well as how the thing works (dynamic, condenser, etc.) but I won't go there to avoid confusing people.  In fact, it would be better if you simply forgot I mentioned that.  (Sorry!)

OK, I think that's more than enough info, I'll stop here...

[Destructavator]

Thought I'd post a little update here,

Some of the shorter and simpler lines are easier to do, but some of the longer ones are rather difficult to make sound convincing.  I've been working over time on more samples, and I must confess it is more difficult than I thought it would be.  There are still a few other problems - I'm just one person meaning one voice, and I haven't found anyone else interested who could use my recording equipment.

Do we still have others interested in doing voices who have made any progress, both male and female?  If not, the talking soldier idea might not work out to well with not enough variety.

I'd really hate to kill the idea, but if we can't get enough voices perhaps we should just work on other sounds that don't actually use words (screams, dying, pain, etc.) for now.

I'm still trying to figure out where I can record a scream, as I live in an apartment building and can't irratate the neighbors.  I suppose I could get in my car and drive out somewhere and reoord it in my car?  (I actually tried that once for one of my songs, I found that it quickly attracted unwanted attention, although I did discover that my scream sounds like a cross between Bruce Willis (the movie actor) and Kirk Hammet (lead guitar player from Metallica).)

[Destructavator]

Wait a sec, I just got an idea:

I've seen lots of GM sample packs for music production that include screams in their SFX kits, many of them very convincing.

Some of these packs are free, distributed under various licenses, I'll do some research into this and post again later if I find any sample packs with a compatible license - I already know of one instrument sample pack that is free with a license saying the author just wants to be emailed if it is used so he knows how many people use it, but the pack is from 2001 and I'm not fully confident the listed email address is still valid.  If it isn't, it would be classified as abandonware, and the legal use of it would be too risky in my opinion.

[XCOMTurcocalypse]

Um, I have a microphone, would it be convenient for me to record a few sentences and send it to you people?

I mean, which ones are the lines that you need? I'm no pro, but I definitely can do some sound. At least screaming or dying in pain...

[michael]

Hi, first post. I love the game a lot, and would be quite pleased if I could contribute.

I've also got a little home-studio, and a decent voice for acting.
Should I just post a flac of a recording of all of those lines?
I could also produce a radio effect, if necessary.

Also, I was thinking, it would be cool to have many different voices for the different sayings, and have which voice is used be specific to each of your soldiers. So, in other words, each soldier will use the same voice consistently, and (most) soldiers will have different voices. I think that would help to give a better identity to your soldiers. Just an idea. =)

[BTAxis]

Ideally, that's what we want. But it's hard enough to even get ONE set of voices, let alone enough different ones for all the soldiers, and then half of them have to be female...

[eleazar]

There are no soldier sounds in trunk.

Has no one actually contributed sounds in all this time, or does someone have them, but hasn't added them to the project?

[BTAxis]

As far as I know, they don't exist.

[eleazar]

OK, here's how i think it could work.

Each vocal file name has the following hierarchy of elements in order, (though "sub event" is optional).  When a soldier speaks, the game will look for a file that matches as many of the elements as possible, and then if there random variations available choose one of those.

This should almost every time choose the most appropriate sound file, assuming multiple sets of sounds many of which are incomplete.


Event -> Gender -> General Voice Type -> Individual Voice -> Sub Event -> Panic Level -> Random Variation


"Panic Level" is from 1 to 5, with "1" being "out of control" and 5 being "very confident".

So a filename would look something like this:

xspotted_m_anglo_eleazar_ortnok_pl5_02.avi

which breaks down to:

Event - Alien spotted
Gender - Male
General Voice Type - Anglo
Individual Voice - Eleazar
Sub Event - specifically the alien is an Ortnok
Panic Level - 5
Random Variation - #2

I'm gonna try to record some.  I have lousy equipment, but sounds will really help the game.

[BTAxis]

I think that's not a good idea. Especially with the variations and panic levels, the amount of samples will be out of control, and it'll take ages to record all of them for just a single voice set. Not to mention that it will put a big strain on the game download size.

[eleazar]

I think that's not a good idea. Especially with the variations and panic levels, the amount of samples will be out of control, and it'll take ages to record all of them for just a single voice set. Not to mention that it will put a big strain on the game download size.

A sound clip saying "roger", "alien spotted", or "i'm hit!" etc. is really quite a small file. Assuming an average length of 2 seconds you can fit 90 such comments in the same number of kilobites an average 3-min song would take up.
I calculate it would take 105 comments to fill out all the talk-circumstances Winter has provided with 5 panic levels and 3 random variations of each one.  Size is not an issue.

But more importantly who cares if it takes ages to record them all?  This plan works even if we never get a "complete" set.  We don't need them all, the filename is hierarchically organized, so the game can easily pick the most appropriate clip from those available.  This approach deals gracefully with incomplete sound sets (a natural circumstance in FOSS).

Besides, i didn't think up the idea of "variations and panic levels"... it's implicit in the quotes Winter has provided.  I've just provided a organized way to deal with them.  It can get really annoying when the soldier says "yes sir!" in precisely the same way every time you tell him to move.  Random variations in sound responses have been part of gaming since (or before) Warcraft... it's not an extravagance.

[BTAxis]

I disagree. I don't think an average sample length of 2 seconds is a realistic assumption, given the example sentences in the first post. Also, I don't think you can realistically include 5 panic levels. It takes a very good voice artist to make the difference even noticeable with that kind of gradation, and I really don't see the added value.

And yes, size IS an issue. Remember that we need as many voice sets as possible, for both genders - and maybe in different languages too! I think you underestimate how much that will affect the download of the game package.

Here's a few figures for you, based on your own assumptions.

105 samples per voice set, averaging 2 seconds each for a total of 210 seconds (3:30). In OGG format, this would be about 10MB (based on my observations from the music folder - see David01.ogg, almost 10MB for 2:41, though there are more efficient ones in there). Let's say we want some 6 different sets for each gender. That gives us 12 sets, meaning 120MB of voices for English ALONE. And as I said, i think the 2 second assumption is too low, plus there might be voice sets in other languages. It gets out of hand real fast.

[eleazar]

105 samples per voice set

105 is the outside, insanely optimistic number of samples.

Also please note: you are using 2 mutually exclusive arguments.
1) there are too many samples, it will take forever to get a complete set
2) we will get too many complete sets, and the DL size will be too big.

In OGG format, this would be about 10MB (based on my observations from the music folder - see David01.ogg, almost 10MB for 2:41, though there are more efficient ones in there)

While we will want to keep a high-quality unprocessed version of the vocals samples with the project, we only need to ship reasonable quality versions (with radio effects perhaps).

I don't use OGG a lot, but it is supposed to be roughly comparable to ACC and MP3.  If so, 10MB per ~3 minutes is audiophile quality, significantly better than what you would get with most downloaded music services, which would weigh in around 1MB per minute.  However the human speaking voice can be compressed smaller than music without noticeable loss of quality.

The professional sounding audiobooks i listen to are generally recorded at 64pbps.  I wipped up a vocal OGG at that rate and it came in at .4 MP per minute.  At this size, the blue-sky scenario of 12 full sets would weigh around 17 MB.


If we ever get to the point of having sound sets in alternate languages, i think they could be downloaded separately, to avoid a game bloated with content most players won't use... but that won't be an issue for some time.

[BTAxis]

Also please note: you are using 2 mutually exclusive arguments.
1) there are too many samples, it will take forever to get a complete set
2) we will get too many complete sets, and the DL size will be too big.

I don't follow; those statements are not mutually exclusive. Unless you took "forever" literally, in which case I must clarify. I mearely meant "a long time".

While we will want to keep a high-quality unprocessed version of the vocals samples with the project, we only need to ship reasonable quality versions (with radio effects perhaps).

I don't use OGG a lot, but it is supposed to be roughly comparable to ACC and MP3.  If so, 10MB per ~3 minutes is audiophile quality, significantly better than what you would get with most downloaded music services, which would weigh in around 1MB per minute.  However the human speaking voice can be compressed smaller than music without noticeable loss of quality

Well, like I said, I based the size on an observation I made on files *we are using right now*. I don't see how that is somehow comparing apples with eggs, as you seem to be implying. As for the quality, reduction is most certainly noticeable, same as with music. You may be referring to the fact that human speech remains understandable even at low quality levels.

I really think we should drop the panic levels because, as I said, I don't see the point in having them.

[eleazar]

I don't see how that is somehow comparing apples with eggs, as you seem to be implying.

The entirety of my last post didn't take.  I edited it to include some hard numbers and other content which you didn't see before making your last post.  However .4 MB per minute vs 10 MB per minute is a big difference, weather or not you consider it "apples to eggs".  (Is that the normal idiom in your language?  In American English we usually say "apples to oranges.")

As for the quality, reduction is most certainly noticeable, same as with music. You may be referring to the fact that human speech remains understandable even at low quality levels.

The best phone call you have ever experienced was at quality rate far below an 10MB per minute OGG.  Was it noticeable?... probably if you tried to notice... but that doesn't mean it detracted from the experience in any way.  Similarly the textures are noticeably pixilated at close zoom, and are models are noticeably made of polygons... is that a fatal problem?  Or in the area of sound also, why shouldn't we make a reasonable compromise between quality and file-size/speed?

I really think we should drop the panic levels because, as I said, I don't see the point in having them.

Simply because it will take a kill the dramatic atmosphere, if a spooked soldier, who is on the verge of homicidal panic... sounds just as in-control as he always does.  Sound is one of the most effective way to convey a mood in a game. People will instinctively believe vocal cues over the numbers.
But feel free to record/acquire samples without different panic levels.  My plan can accommodate that.