Collecting audio

[fangbaby]

Hello everyone, I'm Fangbaby.

Background:
Engineer/programmer/linux nerd/musician.
As a nerd, when I started making music in a band, I pretty much couldn't resist the technical challenges of recording. I have some effective recording equipment and it is somewhat mobile.

Plan:
I've got a car nerd buddy with a Miata, so I figure I'll record an array of sounds and go from there.

In and out of the car versions of:
Engine noise at various RPM
Shifting sound
Tire squeal
horn

Future:
Once we find a set of sounds that works, I'll find more people/cars in the Austin, TX area and record the same set of sounds from other cars.

I'll post back here when I have made progress and start a wiki page with some actual content.

Edit: Started a wiki page here:
http://wiki.vdrift.net/Sound/Audio_Development

[joevenzon]

Good luck!

[fangbaby]

Just bumping to say that I did the recordings and put up a wiki page with the results.

http://wiki.vdrift.net/Sound/Audio_Development


Things of interest I didn't put on the wiki.

Tires were too sticky to squeal, a good thing I guess

I didn't think the actual driving sounds were all that great, mostly just wind noise.

Shifting didn't sound like anything, but maybe I should grab some of the engine slowing from high RPM to low.

Lemme know what you think.

[joevenzon]

Those sounds are great! Nice work. I'll see if I can't get some basic support for multiple RPM sounds hacked in by the weekend.

[joevenzon]

Okay, I've got some basic support in R1867 and added sounds to the M7 car in the data repos R115. A few issues that became apparent:

1) The lower RPM sounds are really quiet. Personally, I think this is great... I like the dynamic range. Others might complain, but for now... leave it alone. :-)

2) Some of the RPM loops don't loop all that well.

3) The revs need to be pitch matched... what I mean is, if you take a 3k sample and speed it up by the ratio of 4000/3000, it should have the same pitch as the 4k sample. Some of the samples actually sound pretty good already (which is a testament to the nice engine and precise recording), but it's probably worth it to go through and do this for all of the sounds. If you don't have any idea what I'm talking about I could try to do it, but my guess is you probably have better sound editing software than I.[/list]

[charlieg]

3) The revs need to be pitch matched... what I mean is, if you take a 3k sample and speed it up by the ratio of 4000/3000, it should have the same pitch as the 4k sample.

That doesn't really make sense -- surely part of the reason for recording the rev sounds at different levels is because it is a sound that does not linearly scale in pitch and volume; 4k revs will not sound like 3k * 4/3.

[fangbaby]

1) Ok, I can easily adjust this whenever, so I'll do as you say and leave it.

2) It was hard to tell what would work, can you tell me which ones loop well so I can try to make the others similar. I guess I should check out the code and hear for myself.

3) I'm not sure about this, but it is easy to test. Which loops sound good when this is done and which ones don't? Maybe the ones that don't work are just poor loopers to begin with.

Fantastic work by the way, that was fast!

[joevenzon]

That doesn't really make sense -- surely part of the reason for recording the rev sounds at different levels is because it is a sound that does not linearly scale in pitch and volume; 4k revs will not sound like 3k * 4/3.

So, the way the game uses the multiple RPM sounds is by crossfading between the nearest two RPM sounds. So if you're at 3500k, you'll get 50% gain on the 3k sample (pitched up by 3500/3000) and 50% gain on the 4k sample (pitch down 3500/4000). Unfortunately, if the 4k sample is actually 4050 rpm, then when you're at 3500 playing the 3k pitched up and 4k pitched down at the same time at what should be the same pitch, the 4k sample actually sounds a little higher, and so they don't mix well. It ends up sounding like two engines. Load up the M7, rev to 5500, and you'll hear what I mean.

[joevenzon]

2) It was hard to tell what would work, can you tell me which ones loop well so I can try to make the others similar. I guess I should check out the code and hear for myself.

2k looped very well, 3k looped the worst. The reason that 3k didn't loop very well is because the engine was slightly revving up during the middle of the recording.

[joevenzon]

3) I'm not sure about this, but it is easy to test. Which loops sound good when this is done and which ones don't? Maybe the ones that don't work are just poor loopers to begin with.

I went through and "tuned" the engine samples. I realized I could do it without modifying the sounds themselves; I just changed the RPMs they represent in the .car file. The way I did it was like you'd tune a guitar... I took the 1k sample, doubled it, then adjusted the speed of the 2k sample until they sounded correct when played over each other. Once I knew the speed I converted that back to RPM and put it in the file.

[fangbaby]

I grabbed the sources and data from svn, compiled and checked out the M7. I think it sounds great! I tried another car and it sounded synthetic and fake to me.

Do you want more? Since the external sounds seemed to be best, I could use my better equipment to record it and make it all sound just a little more real.

Should I?
Any other requests?

[joevenzon]

If you're able to grab so road noise sounds, that'd be great. You know, just the ambient, steady sound of moving forward.

I'll try to add in the interior sounds and the engine start sound into the code pretty soon.

[kidrock]

try using the handbrake or heavy gas while turning to get SOME squealing... these are very good though!

[fangbaby]

Updated the wiki with 2 more samples.

I grabbed a couple of spots from the original recording when we were moving and the RPMs were not all over the place. I think it just sounds like random noise, but maybe your magic code will make it sound good.

I will get the squealing sounds, lets see how this works first.

[kidrock]

kk... good to know.