next »

[Characterlimi]

As a freelance composer/sound designer I have had an opportunity to work on many different kinds of projects, and I have experienced many different kinds of audio pipelines. Sometimes you get the opportunity to collaborate with other sound designers or composers, which brings to light interesting problems if you and your collabs are working remotely. The primary problem I want to address is that of dynamic range, and loudness standards (or the lack thereof).

I have seen this discussed in other areas, such as AES meetings, however the general consensus with regards to loudness standards seems to be a resounding "Huh? Wh...what?" followed by a lecture from a veteran (see: old) audio engineer describing the abuse of aggressive compression and the degradation of the art of Mastering. This isn't to be dismissive of the wiser (older) audio guys. I agree with the bulk of their rants. My issue is that nobody seems willing or interested in naming a standard.

This may all seem like an incredibly narrow/niche issue, but if every audio asset coming in is a wildly different level, batch processing becomes a thing of the past. Productivity falls to a stand still because somebody at some point in the pipeline has to sit down and squash or boost everybody else's work. Not to mention that you'll be risking a myriad of distortion issues when you start boosting recorded audio (foley or vox).

I'm not saying I want a be all end all for these issues. I'm fully aware that every project has its own unique set of needs and will have its own set of audio thresholds (a horror might have a much wider dynamic range over all as opposed to a shooter that might feature constant high-octane audio feedback).
 
I guess I just want to start a discussion on this to see if we can at least start a general guideline towards how we handle our audio.

[ProgramGamer]

Considering that sounds and music are such an integral part of game feel, I agree with you 100% on this. As to name a standard though, I have no experience in the field of sound design, so I can't really help you much. My educated guess though would be that you'd want to record everything as loud as humanly possible in order to avoid distortion, since the way I understand it reducing the volume of something doesn't destroy the recording as much as increasing it. I'm also guessing that the eventual standard will be measured in decibels?

Anyways, that's my two cents. Hoping more people will hop on the discussion.

[Lauchsuppe]

Well, you can't really measure perceived loudness. So I guess that's why there are no standards. I guess there's no alternative to having someone in the process master everything. Just make the individual audio assets as loud as possible (w/o additional compression).
Luckily, games don't need to directly compare themselves to other games in terms of audio volume - that means it's not that much of a problem if a particular game is not as loud as others. "loudness war" doesn't really need to be a thing as much as it is a thing in popmusic. Actually, I think you can make out huge differences in terms of overall loudness even in AAA games. I can't give any specific examples right now and it's more a gut feeling - but I remember having to turn up the volume quite a bit for some particular games.

[rmrzero4]

Mixing and mastering is ultimately a matter of personal preference. Also there's the variable of perception; everyone hears differently. I am also quite certain this is why there's no loudness standard. As a general rule of thumb... I was always told to just watch levels and make sure you have enough room to go loud if you're starting off soft, and it's always a good idea to not be lazy and mix and master at the end all over again. Heavy use of compressors and limiters were discouraged in my university however they are handy tools, as "destructive" as they are to the audio quality.

When it comes down to it, I just mix and master how I see fit, and then ask someone else. It'd help to ask someone who knows nothing about audio, and someone who does. Overall, your audience isn't always going to know audio so as long as they enjoy your mix you should be fine.

[oahda]

This annoys me a lot too. There are standard resolutions but not standard audio volumes; I've noticed this especially flipping from TV (I don't watch TV but I have to start the TV set to a channel and then flip to AV or w/e for games/movies) to movies/games there is such an enormous difference (TV is always a lot louder) and the same goes for moving between YT vids and so on.

A lot of programs have an equalisation option, tho. Doesn't that generally bring about an even (tho very loud but you can always decrease volume of things too loud whereas if the sound is too low even on full volume you're screwed) level, similar no matter what software or sound file you use? If so, that'd be an option, I guess, hoping that at least some others might use that way. .-. But I'm not an audio person so you'll have to tell me whether I've misunderstood equalisation.

[Lauchsuppe]

well, there is a "standard" audio volume which would be 0db for digitalized audio. this is the highest possible loudness. if you exceed this by using an equalizer, you'll end up with distortion. generally, how the human ear perveices loudness is subjective and a very complicated thing. for example, the human ear is esp. sensitive in the frequency range of human speech. so even though two audio signals peak at 0db, one of them might actually be perceived louder than the other one.

also, if a piece of audio would peak at 0db all the time it would be super dull. so there's this thing called "dynamic range". dynamic range describes the (perceived) loudness difference between the quietest and the loudest sounds within an audio track. depending on the kind of audio you make, you may want to have a high dynamic range: for example in horror games, where you may want to have all these small details floating around in the soundscape to stimulate tension. and then - for a jumpscare situation - you might want to drop an audio cue significantly louder than the rest of the audio scenery. in order for this to work, the rest of the game needs to be significally quieter than 0db.

the reason why tv shows or popsongs on the radio are often way louder is because they're often optimized to be as loud as possible. that's becaise usually, the more loud a piece of music is, the more intense and better it is perceived by the listener. a recent pop single that is intended for airplay would usually have a rather low dynamic range in order to be as loud as possible all the time. on the other hand, a classical radio station is usually less loud because a symphonic orchestra traditionally needs a higher dynamic range in order to work properly. bc symphonic orchestras were invented for concert halls, not airplay.

so it's always about finding a compromise between loudness and dynamic range. that's why it's not standardized. instead, there's specialized people for adjusting audio so it fits in with other cues/tracks called mastering engineers. but ofc it's nowhere as reliable as standards

[oahda]

Doesn't the volume controller on the device turn the decibels up and so introduce distortion anyway if I think something is too low at 0 and turn it up? Or is 0db the loudest setting for any speaker?

I'm thinking that there could still be a standard "normal volume" that people work around when they want to add the effects you're talking about. So that if I turn the TV on and nothing special is happening and then swap to a movie where nothing special is happening I won't have to turn the volume up a lot because they didn't have the same general volume? If the TV set is not on full volume, would >0db according to the sound file actually be played at >0db from the speakers? If not, then there wouldn't be a problem, would it not?

Agh. I hope you're understanding my thinking. There must be some way to normalise or generalise this because the current situation is really annoying not only for makers but especially for users who have to keep adjusting the volume all the time. ;_;

[Lauchsuppe]

0db is just the digitally highest peak. that's why you can't turn up volume above 100% in the windows speaker settings for example. while in vlc media player - which let's you set volume to 200% - you'll notice distortion if you go above 100%. your speakers can amplify the audio volume significally because at that point the audio signal is analog already and not digital anymore. I'm probably not knowledgable enough to explain this more in depth ; but for physical analog audio signals there is of course no limit. the digital 0db measurement is just a concept; like a scale. 0db is loudest signal and -inf db is quietest (=silent) signal. your pc speaker's job is it to amplify this concept to an audible level you like.
it's a little irritating, but the 0db to -inf db concept has nothing to do with how dezibels are used to measure actual real world audio signals.

I don't think it's possible to find a standard volume to work around because people would always want to have their audio to be as loud as possible without losing dynamic quality. also, like I said, perceived loudness is nothing you could really measure. and perceived loudness can be very different from actual physical loudness. however, mastering tools and processes are getting better and better. maybe in the future tools will allow to have a symphonic orchestra to be played as loud as a modern popsong without it losing it's qualities.

you can check this out if you want to read more on it, that article explains it better than I could https://en.wikipedia.org/wiki/Loudness_war

[Characterlimi]

I honestly wasn't expecting a conversation to blossom from my post, so first off, thanks to everybody who's contributed so far.

Lauchsuppe is correct in stating that 0dB is the peak at full scale in a digital system and that the perception of an audio source's loudness is a subjective experience.

I don't want to rehash several courses of audio engineering in this thread, but suffice to say that the fletcher-munson curve https://en.wikipedia.org/wiki/Fletcher%E2%80%93Munson_curves will go a long way in explaining which frequencies humans tend to be more sensitive to.

Again I'm not looking for an end all be all, I just think it's important for people to think about these issues. I've always believed that if there is an opportunity to increase efficiency then it should be taken. And don't get me wrong, I firmly believe that certain games need assets of varying sound levels, theres a reason you spend hours tweaking the mix. I guess my original post was just to vent frustrations with having to handle other peoples audio assets, especially after any amount of DSP is applied to the assets. Maybe I should just grow up and spend more time in the mix 

In the same vein, I know there are conceptual mix solutions out there that have been dreamed up by some of the smartest guys in the industry right now, for example the HDR (High Dynamic Range) by the guys at DICE and PDR (Parametric Dynamic Range) by the team over at Naughty Dog. What do you guys think of these systems? Are my initial complaints invalidated by emerging technologies?

[oahda]

Geez, and I thought HDR was related to video (and like ten years older than Dice for that matter...).

[bbtombo]

I think there are two things to consider. For any sort of industry standard, the mixing and mastering needs to be well done, even in a basic form. That can go as simply as "no peaking" when it comes to overall volume. I think it's also important to respect how it sounds with the sound effects of the project! "Loudness" is determined by the user in the sound settings, so it should just overall sound "good."

[TopherPirkl]

Incidentally, there actually is a standard that's been bouncing around for a while, published in a paper (http://gameaudiopodcast.com/ASWG-R001.pdf) by Sony's Audio Standards Working Group. It was specifically developed for Sony's systems, but a lot of developers are using it as a general guideline for loudness standards on any platform.

In short, its recommendations are that you meter the True Peak output of the platform for a minimum of 30 minutes to record an average LUFS (article uses LKFS, but they're identical and it seems LUFS is becoming the more common descriptor) of -24 LUFS for consoles and -18 LUFS for portable, with a 2 LUFS margin. Obviously the gaping hole is PCs, since the sound card is a major determinator of the actual LUFS at output, and you can't assume a configuration like you can on a console/Vita.

Also, the paper recommends that no individual peak exceeds -1 LUFS, which I'd second. Realistically, if you're creating assets/music, I'd recommend never tripping -1dBFS, and never normalizing beyond this. You'll see -0.3 or -0.1 thrown around a lot, as people want assets as loud as possible so they can attenuate them in the game engine, rather than adding gain in-engine, but with anything about -1 you run the risk of having intersample peaks when it's converted back to audio, especially on consumer-grade sound cards.

I've used Orban Loudness Meter in the past to track this kind of thing. Beyond that, I couldn't say much about it, I'm an audio engineer in job title only, I don't actually have in-depth engineering education.


As for HDR systems, I've only ever used Wwise's HDR, but it's amazing how much time it saves over using a complex ducking structure. I think the thing that people tend to forget is that you can use it for creative purposes, rather than just raw volume control. For example, doing things like ducking all audio, even for just a brief handful of milliseconds, while a player's ultimate attack lands, can add a ton of impact and make that ultimate seem much more loud and impactful than it really is. Destiny does this with the players' ultras; it fades and mutes everything long enough for the attack's loudest transient to land, then fades it all back in, pretty quickly. (Ironically, though Destiny uses Wwise for its audio system, it actually doesn't use Wwise's HDR system. I think it predated Wwise HDR when it first went into development. )

[Characterlimi]

TopherPirkl, all of the words you have said are words that I wanted, thank you