Hacker News Reader

Comment by owenversteeg | Hacker News Reader

So, semi-related question which I'm sure somebody here on HN knows the answer:

I've been wanting to make a programmatic car "vroom" noise. As in, you can arbitrarily change the speed/acceleration and there are realistic sounding engine noises. Acceleration, deceleration, constant speed, etc. I've thought of making this by stitching together a bunch of car noises, but all my attempts have ended really badly (partly because most car noises are relatively low quality and partly because they don't stitch together well.)

So how can I make realistic engine noise? I'd appreciate anything that might help (recordings, links, examples of it done, etc.)

View on HN

I will once again link to xoxos VST plugins (my fanboyism is showing):

http://www.xoxos.net/vst/vst.html#models

Scroll down to the "virtual machine" plugin. It's closed source but there is a PDF that explains how it works, with a nice signal flow graph. You can also get a very good idea of the underlying algorithm by playing around with the parameters.

Also, since someone else is asking about generating squeaking tire sounds, xoxos also has a "friction" VST that does this. There's no audio demo so I made a quick one that emulates tire squeaking:

http://007ee821dfb24ea1133d-f5304285da51469c5fdbbb05c1bdfa60...

Edit: and while I'm at it, here's another engine sound demo using "virtual machine" (less interesting than the one on his website, but I put a bit of reverb/ambience on it to make it more realistic):

http://007ee821dfb24ea1133d-f5304285da51469c5fdbbb05c1bdfa60...

Those attached PDFs give a great overview of how to implement this. The virtual machine and friction plugins sound so good, I really want to try duplicating this. Thanks for the samples!

Star Citizen's recent ship production pipeline video had an overview of their thruster audio. It's not a perfect reference since they're not trying to reproduce a real-world sound, but it might be of interest: https://youtu.be/Bbq2Crt8pTA?t=1525

Segment on engines starts at 26:44.

EDIT: for anyone who can't/won't watch the video, I think it's a bunch of synth and sample layers that shift at different rates as the thrust changes. He only specifically talks about a few synth layers though.

This seems like the sort of thing you could do procedurally relatively easily, without recordings. Find some car recordings and stick them in a spectral analyzer like Raven. You'll probably see harmonics of the valve and crankshaft periods, with intensity changing with engine power.

That sounds like something you should be able to do with subtractive synthesis. If you don't have an analog synthesizer lying around, you can try software like maybe zynAddSubFx. If you set the pitch bend range wide, you could simulate smoothly revving from low to high RPM.

I have a Moog Slim Phatty that does an interesting/glitchy thing if you try to play notes that are too low for it, where at some threshold the oscillators slow way down. If you use pitch bend to ride the edge of that threshold, it sounds like a car engine trying to turn over, catching for a few cycles, stalling, and so on.

I'm pretty sure that videogames do this by using lots of samples that they combine to have variation, same thing that realistic drum samplers do, they take hundreds of samples and combine them.

I've been struggling to synthesize satisfying engine sound and tire squeal for a game, my best efforts sound like an engine that runs on soap bubbles. I think the cutting edge solution involves harmonic analysis of samples to find or make smooth looping points within the engine cycles. Crankcase audio has some tool that does this https://www.youtube.com/watch?v=kpeJdErx5b4 . Way too expensive for me though.

Physical modeling / waveguides / etc. might be one way to go here.

As an example: Andy Farnell's "Designing Sound" (https://mitpress.mit.edu/books/designing-sound) describes some physical modeling techniques for engines. (The model is referenced in this paper: http://www.academia.edu/14107512/Physically_informed_car_eng... which tries to improve on the model and also might contain some additional ideas).

I took a listen at the engine sample from Andy's PureData code page (http://aspress.co.uk/sd/practical22.html). The second half (where the engine was revving and accelerating) sounded pretty good to me! It might need some combination with samples, though, the idling portion kind of sounded off...

This is really cool, modeling the process instead of relying on samples is a really appealing solution. The first engine sounded like atari pole position to me but the second 4 stroke model is pretty amazing. I guess the added complexity makes the engine sound more appealing.

Reply on news.ycombinator.com