So, @Mangmod made a little bit of researches about some realistic sound effects in this video and they seems quite better, but some of these sounds has a little bit of people talking, but still realistic:
Maybe they can try to remove some vocals and background noises using AI.
Yes, there are some noises in the background. Please keep in mind I added artificial delay and reverb, as well as a background noise, or the sounds would be very dry. FL Studio has now AI tools to separate vocals so I might run the videos I have through it.
Here’s a video with the DRY sounds. As you can hear it’s much less realistic without the noise.
Nevertheless, let me use this thread to explain a proposal for FWSim’s audio engine.
The software could use sound blending and filtering depending on distance. Here’s a video demo which could help understand what I mean:
It would work like this:
There would be three layers:
– Nearfield - which is incredibly close to the mortar (danger zone)
– Medium - here mortars are heard in their transients. About 150m from the area of firing
– Spectator - the core sounds, which should hypothetically be the ones to hear as a spectator in a distance of about 400+ meters
At very close range, play only layer 1 (nearfield)
After 50m, start blending layer 2 with layer 1 (basically progressively reduce volume of layer 1 and increase the volume of layer 2 with the increase of distance)
After 150m, start blending layer 2 with layer 3.
After 500m, start adding equalization on layer 3, to “muffle” the sounds and remove some frequencies - or the opposite -
There are some notes to be made, however, mainly:
There would be a lot of work to do, code-wise. I am not a coder nor an audio engineer so I cannot make any estimation on this. I suppose there are some free libraries which could be used to implement a similar audio engine.
There would be some sound to be made. Not only mortars and explosions but as @fy1vo51 pointed out, crackles, crossettes, popping flowers, whistles, fountains and the various howling effects.
The sounds should be consistent with the calibre, so perhaps they would need to be divided into the possible categories:
– Mines/Comets/Ground effects: as I could hear in real life, mines and comet launches are much louder than what you hear in FWSim. Even at a certain distance, they can be heard distinctively.
– Micro: consumer cakes mainly, from 15 to 50mm
– Small: anything from 50 to about 150
– Medium: 151 to 250mm
– Large: 251mm+
I wouldn’t add an extra large category due to the sheer difficulty in finding sounds of that calibre or editing them enough to get a satisfying result.
Speed of sound should be taken in as well to guarantee a more realistic result (could be a toggle)
As an alternative to this system, have a “sound libraries collection” from which the user can choose which effects to replace with so you can spam the 500mm sound on the small shells too
Implement a Convolution Reverb system, which is a “simple” way to get a realistic reverb hopefully without impacting performance much. This would let you select an impulse response from a list and use it as the reverb sound.
With this in mind, here’s the proposed pipeline:
Get the distance of firing
Depending on the distance, randomly pick one or more sounds from the pool depending on the type/calibre of the firework
Depending on the distance, blend the sounds, based on their type
Create a micro-pitch variation to add variety to the sound
Depending on the distance, determine the time to play the sound (speed of sound simulation)
Once you find all sounds, you can try to add all effects (reverb (5-10 seconds), vocals and background noises removed) to every sound files and then send it a packed file so I can review them and send to Lukas.
Thanks for your thoughts and experiments. As some of you may know, we made an improvement to our sounds back in 2020. Unfortunately, we didn’t do a great job and some sounds got worse.
We are interested to re-do the sounds again in 2024, however it is unclear if we will actually have time to do that.
FWsim is using an audio system called Fmod. Fmod is great because it allows sound designers to modify the sounds without needing to write code. If some people would like to try their hand at improving the sounds, we might be able to provide the Fmod project files. You could load the existing FWsim audio system into Fmod and then try replacing the sounds with new ones, and see if you can come up with something good.
If you or other people are interested in receiving those Fmod project files, please post in this thread. If there is enough interest we might do this (no promises though).
This would be a great opportunity to try something practical. I would like to see how the system works and eventually learn it so that sound replacements/additions can be made. It would be handy to actually test it with a smaller batch of sounds rather than making all of them and constantly editing them in the future.