justin = { main feed , music , code , askjf , pubkey };
Ask Justin Frankel
No reasonable question unanswered since 2009!

Suggested topics: programming, music, sleep, coffee, etc.

Note: please do not ask questions about REAPER features, bugs or scheduling, use the forums instead.


Name: Ask: Human (enter yes):
[unreplied] | [replied] | [recent comments] | [all]
[older questions]

Recent comments:
  • Posted by Xavier (92.153.75.x) on June 17 2026, 9:29am in 8175:
    Thank you!

  • Posted by gfx accuracy (102.176.65.x) on June 12 2026, 12:44pm in 8173:
    Great, will take a look at it. Thanks

  • Posted by Matthew (98.13.80.x) on June 8 2026, 2:23am in 8166:
    I have a feeling it may have to do with it be being a very optimized piece of software. While I am not a developer myself of Reaper. Coming from a developer perspective and software engineering in general. I am very confident that Reaper is heavily tested and refactored in order to meet efficiency margins. But again, that's my speculation. But regardless, Reaper is a very solid piece of software.

  • Posted by atmo (93.89.128.x) on June 7 2026, 5:00pm in 8154:
    your story about debugging the chromebook made me wonder - how is your google-fu these days now that everything is ai generated blogspam? what's the best way to find real, useful material on the internet still? I saw Jon Gruber from daringfireball (creator of markdown) giving a strong endorsement for Kagi, which is a paid search engine. not been desperate enough to try it yet.

  • Posted by robert (92.157.64.x) on June 6 2026, 6:35am in 8160:
    These guys just want to make money. They'll be more interested in selling what they've discovered than keeping it for themselves.

  • Posted by Sunblister (77.187.11.x) on June 5 2026, 6:53am in 8160:
    According to The Banality of Evil, them being human is far worse than monsters. You shouldn't see everything in such an overly positive light.

  • Posted by Hahaha (154.161.244.x) on June 4 2026, 10:19pm in 8167:
    Yeah, I agree. Lol

  • Posted by robert (92.157.64.x) on June 4 2026, 11:30am in 8160:
    You shouldn't see everything in such a negative light. The leaders of Silicon Valley aren't monsters.

  • Posted by Xavier (128.79.126.x) on June 3 2026, 7:49am in 8160:
    Those are 100% ethics-driven, trustworthy people. And, should such a claim ever become real, they would gladly share this longevity pill with all of Humanity for free, with no slavery involved. Oh, and did you hear that, thanks to AI, we won't have to work anymore because we'll all have Universal Basic Income? Oh, my sweet Summer child.

  • Posted by robert (92.157.64.x) on June 2 2026, 4:50pm in 8160:
    Every Silicon Valley tech leader claims that we will be able to live to at least 150 years old before 2035.

  • Posted by Xavier (128.79.126.x) on June 2 2026, 10:04am in 8160:
    "Once superintelligence is here and allows us to live to be 150", that's a good one :) If AGI ever happens (and I doubt it ever will), you can be sure than it won't be used to improve humankind in any way, shape, or form. As to living to be 150, first, won't happen; second, who would want that?; and third, none of us will be rich enough to pay for longer life.

  • Posted by ext-tail-size-1  (154.161.251.x) on June 1 2026, 7:37pm in 8165:
    Thanks so much. I now understand.

  • Posted by Justin on June 1 2026, 7:05pm in 8165:
    ah sorry yes, I was mistaken. you can use 0 (the default, the same as not setting it) which means "can produce output at any time", or -1 which will auto-detect silence.

  • Posted by ext-tail-size-1  (154.161.254.x) on June 1 2026, 6:41pm in 8165:
    The difference between 0 -1 -2 and not writing it all is confusing me. If you could explain on the context of high Q and delays vs plugins that simple like only a gain knob. I want to know when to ignore writing ext-tail-size and when to use -1 0 or -2

  • Posted by ext-tail-size-1  (154.161.254.x) on June 1 2026, 6:37pm in 8165:
    Also, what happens when ext-tail-size-1  is not written in the code at all? Is that the same as setting it to 0.

  • Posted by ext-tail-size-1  (154.161.254.x) on June 1 2026, 6:27pm in 8165:
    I saw the ff in the manual and I thought -1 means it will detect the tail by itself, but u just said -1 means no tail. "If set to -1, REAPER will use automatic output silence detection and let plug-in state settle. If set to -2, then REAPER will assume the plug-in has no tail and no inter-sample state." So I guess I shouldn't use it all even in plugin with high Q Factor and even in delay plugins ???

  • Posted by wasereb4 (176.1.58.x) on May 31 2026, 8:59pm in 8162:
    youtube.com/watch?v=jLISvFunIFk

  • Posted by robert (88.167.123.x) on May 31 2026, 5:10pm in 8160:
    i m not a troll :)

  • Posted by Sunblister (147.161.164.x) on May 30 2026, 6:16pm in 8160:
    Not sure if robert is trolling, but I think once the AI bubble is popped and the freemium models aren't subsidised anymore, people weigh actually learning something themselves against the cost of tokens per prompt. Sad that the motivation doesn't seem to come intrinsically, but I bet the madness doesn't hold on for too long anymore.

  • Posted by bistro (31.37.8.x) on May 30 2026, 7:21am in 8160:
    i guess the ai also can't know exactly how to implement this or that because choices have to be made. part of why i like repaer so much is because of the millions of choices (because devs are users) that make the program nice to use. jokingly id say that justin is not a developper for reaper but a "maintainer" of a tool and the "context" is huge (is it annoying to use, etc). i'm being vague sorry

  • Posted by robert (92.157.64.x) on May 29 2026, 6:00am in 8160:
    Once superintelligence is here and allows us to live to be 150, things might be different...

  • Posted by Justin on May 28 2026, 10:18pm in 8160:
    (sorry that was cheeky) -- but the hard part isn't the programming, it's the programming things _correctly_ and a lot of that is stuff that LLMs can't do. it's one thing to spit out a bunch of code that integrates a lot of libraries and mostly works, but making something like a DAW is a different beast entirely.

  • Posted by Justin on May 28 2026, 10:17pm in 8160:
    keep dreaming

  • Posted by robert (92.157.64.x) on May 28 2026, 7:43pm in 8160:
    Yeah, but it would help you work and finalize those annoying, ongoing things to settle them once and for all: MIDI editor, the score editor, etc...

  • Posted by Justin on May 28 2026, 1:03am in 8159:
    yeah heaps are usually fast but the concurrency issues are there. maybe you have arrays of structs / structs of arrays backwards there, most people start with arrays of structs...

  • Posted by Justin on May 28 2026, 1:02am in 8160:
    even if that does happen, then all of the big facebooks of the world will fuck us all over by burning tokens to copy anything good. but it's moot, it will continue to be shitty. just less shitty.

  • Posted by jack (91.90.123.x) on May 27 2026, 10:09pm in 8159:
    re:games You rarely need data structures more complicated than simple arrays, try and stay cache friendly, switch to arrays of structs rather than structs of arrays depending on access patterns, fixed-size stuff as much as you can when bounds are known, and allocate all resources up front, manage it yourself, no runtime allocations. Those are the big ones from my experience.

  • Posted by robert (92.157.64.x) on May 27 2026, 9:41pm in 8160:
    Maybe in 6 months they won't make shitty code anymore :)

  • Posted by Justin on May 27 2026, 8:17pm in 8160:
    we'll see! AIs can make lots of shitty code, that's for sure, but it's still shitty code.

  • Posted by robert (92.157.64.x) on May 27 2026, 7:17pm in 8160:
    You say that now, but in 6 months you’ll probably be using AI like everyone else ;) Otherwise, Cockos will be replaced by AIs making music software on demand, implementing whatever people ask them to :)

  • Posted by Mespotine (82.135.127.x) on May 27 2026, 6:02pm in 8157:
    Oh nice, will check. Thnx 🙂

  • Posted by Justin on May 26 2026, 11:23pm in 8158:
    Sounds extremely plausible and likely!

  • Posted by dopp (64.176.64.x) on May 26 2026, 11:17pm in 8158:
    I just remember you were saying that you would like to vote for him, but because he will not get enough votes, you will vote for some other party. Today i see why you liked him, you were right back then, so i was just curious if you still think that way. Thanks for answer.

  • Posted by Stephen Tack (97.120.249.x) on May 26 2026, 1:08am in 8154:
    In a world of disposable Chromebooks, I love that you put the effort in to keep it useful. 👍

  • Posted by init vs slider (154.161.227.x) on May 25 2026, 4:37pm in 8156:
    Alright, thanks

  • Posted by Justin on May 24 2026, 4:29pm in 8155:
    ah good point, hadn’t thought about that

  • Posted by Sunblister (77.187.145.x) on May 24 2026, 11:01am in 8155:
    While I don't know the format, I think that there's also an accessibility side to it where it's less about appeal but more about perceiving the lyrics at all.

  • Posted by atmo (93.89.128.x) on May 20 2026, 2:39pm in 8150:
    cool! thanks for answering

  • Posted by Justin on May 20 2026, 1:39pm in 8150:
    ah -- getting the video frame that is being displayed would be a lot easier actually! let me look into that.

  • Posted by atmo (93.89.128.x) on May 20 2026, 1:12pm in 8150:
    Good question! Hadn't thought that far tbh. I could see use cases for individual media items or tracks (like when I was making a multicam view/editor) but was thinking more along the lines of the mixdown at the end of the video pipeline actually, ie what's shown in the preview window. Maybe GetAudioAccesorSamples was a bad example to give. This thread prompted me to ask - forum.cockos.com/showthread.php?t=281653- although I don't personally have any use for Blackmagic Decklink output myself.

  • Posted by Justin on May 13 2026, 10:39pm in 8147:
    hard to know based on your context, but for example writing the web interface for reaper, where you have a c++ app that does something, a small web server, and then some html/javascript that runs on that... could be a lot of things though?

  • Posted by Megaman (223.185.24.x) on May 13 2026, 6:53pm in 8147:
    Can you give an example of such full stack dev? Or how/what it will look like?

  • Posted by Opinion (154.161.125.x) on May 13 2026, 12:09pm in 8137:
    I really appreciate you talking time to answer. Great. also read your article on it. All these explanations and the kvr debate can be a new article too. Thanks.

  • Posted by Andrey (109.106.143.x) on May 12 2026, 3:39pm in 8139:
    Thank you very much. Add it to Reaper, I think it's necessary for many people who work with RS5k.

  • Posted by Justin on May 12 2026, 2:01pm in 8139:
    ok updated 1014.org/shiz/choke_group2.txt

  • Posted by Justin on May 12 2026, 1:58pm in 8137:
    "there is no worldwide law on this matter of pure AI code being automatically public domain?" correct that there is not much worldwide law regarding copyright in general

  • Posted by Justin on May 12 2026, 1:54pm in 8137:
    "AI-generated code for a software, whether modified or not, can be presented as a human work and receive copyright protection, right? If yes, then the whole thing now boils down to honesty on the part of the developers, right?" potentially, though if it were to be litigated, you could get an expert to show that the code was AI-generated and thus not copyrightable...

  • Posted by Justin on May 12 2026, 1:53pm in 8137:
    "the developer is obliged to release the pure AI code to the public?" they are not, there is no obligation as such

  • Posted by Justin on May 12 2026, 1:52pm in 8137:
    "later added pure AI code, how is the license of the whole product going to be now" doesn't matter -- you can incorporate public domain into software without the whole thing being public domain.

  • Posted by Kevin (142.188.209.x) on May 11 2026, 7:45pm in 8138:
    nail the web/backend and see if the community jumps in creating apps - some of the best navidrome and audiobookshelf apps are 3rd party.

  • Posted by Andrey (109.106.143.x) on May 11 2026, 7:05pm in 8139:
    It works, but because note-offs are enabled, you need to hold down the key for the sample to play fully. Is it possible to make a sample play without holding down a key?

  • Posted by Opinion (154.161.226.x) on May 11 2026, 6:08pm in 8137:
    Lastly, as you mentioned, no one can really know for certain whether a code is purely AI-generated or AI-generated with human modifications, right? So that means AI-generated code for a software, whether modified or not, can be presented as a human work and receive copyright protection, right? If yes, then the whole thing now boils down to honesty on the part of the developers, right?

  • Posted by Opinion (154.161.226.x) on May 11 2026, 5:56pm in 8137:
    Also, if you have released a closed-source software like REAPER, and later added pure AI code, how is the license of the whole product going to be now? Because software usually come with one license. So the developer is obliged to release the pure AI code to the public??

  • Posted by Opinion (154.161.226.x) on May 11 2026, 5:54pm in 8137:
    Great. So I'm summary: 1. Pure AI code (public domain) 2. Edited AI code (edited part copyrightable) 3. Pure AI code used as a reference and re-written entirely by human (copyrightable). So is the first point a USA thing, as one was arguing that there is no worldwide law on this matter of pure AI code being automatically public domain?

  • Posted by whipped-ass-llama (46.164.114.x) on May 10 2026, 7:35am in 8136:
    I agree the same, I visit here time by time and read his opinions.

  • Posted by Matthew (98.13.80.x) on May 10 2026, 1:06am in 8134:
    True. I just like XML for it's verbosity and explicitness. And I will look into JSON5. Thanks!

  • Posted by Justin on May 9 2026, 4:48pm in 8134:
    not gonna lie XML or JSON are both better than RDF/Turtle lol

  • Posted by Tony (125.237.124.x) on May 8 2026, 9:51pm in 8134:
    Maybe I need to get out more, but I haven't heard anyone say "I prefer XML" for a very long time :-)

  • Posted by _progboii (189.7.87.x) on May 6 2026, 12:18am in 7781:
    Hey Justin. I must confess that the fact you said chances of acquiring the extension were close to zero, made me pretty upset. But not your fault, of course! Anyways... Would you at least be willing to take a look at it and see if it would be of any use? Otherwise I think I'll just turn it into my own product without Reaper (it's not a direct competition though!)

  • Posted by Justin on May 4 2026, 8:50pm in 8120:
    I was theorizing but there might be other limits I'm not thinking about.

  • Posted by Justin on May 4 2026, 8:49pm in 8120:
    entirely possible it's still limited to 60 years, or even a month, heh.

  • Posted by Mespotine (212.122.61.x) on May 4 2026, 8:33am in 8120:
    I once tried and Reaper crashed when the project was around 60+ years long. But don't know if you have expanded it in the meantime...

  • Posted by wasereb4 (31.16.104.x) on May 3 2026, 4:17am in 8120:
    Indeed. Wouldn't have guessed *that* long.

  • Posted by Justin on May 2 2026, 10:34pm in 8118:
    WDL is more a set of helper stuff, less frameworky. The crossplatform UI is done via SWELL (which emulates win32 on other OSes) plus LICE (an image drawing thing) plus WDL_VirtualWnd for controls, etc.

  • Posted by Kevin (152.173.82.x) on May 2 2026, 10:05pm in 8118:
    its*

  • Posted by Kevin (152.173.82.x) on May 2 2026, 10:04pm in 8118:
    Doesn't REAPER use it's own thing, WDL/whittle?

  • Posted by Andreas (212.84.135.x) on April 29 2026, 8:46am in 8106:
    You can take a look at CLAP (cleveraudio.org/. It is way less complicated than VST, tho still not as simple as raylib. Not all DAWs support it but you can wrap it as VST (github.com/free-audio/clap-wrapper

  • Posted by Justin on April 28 2026, 4:55pm in 8111:
    yeah probably, attackers tend to not target individual sites, instead writing bots that target common frameworks/products, because they can write once and reach millions of sites. if you make something unique, nobody would bother. and when they do, you can respond by improving it...

  • Posted by blahblih (31.37.8.x) on April 28 2026, 1:18pm in 8111:
    You mean there are not enough visitors to interest attackers? What kind of protection would you have to set up? I'm interested because I'm setting up a blog with (moderated) comments for someone and I'm very afraid of security/AI issues...

  • Posted by jason (78.243.35.x) on April 27 2026, 6:30pm in 8110:
    i think AI will degrade 90% of all production due to poor implementation. Beware of the unintended consequences.

  • Posted by wasereb4 (178.24.51.x) on April 26 2026, 4:44am in 8108:
    Now this I found actually funny (got the reference) lol.

  • Posted by Wyatt (66.171.94.x) on April 24 2026, 7:12pm in 8102:
    Wow. Thanks so much. I like the newer vista and above folder dialog. If a user is still on XP, does it fall back to the older dialog?

  • Posted by Justin on April 24 2026, 12:27pm in 8110:
    there's a "report file" thing.. but really only for spam/malware/non-reaper-related stuff, we don't want to be censoring too much, maybe some of the slop is useful?

  • Posted by omitz (31.37.8.x) on April 24 2026, 11:00am in 8110:
    I beg you Justin, keep the basic version of reaper just as it is, invulnerable from such problems! (even though I admit I'm very glad you used ai to make ffmpeg work :p)

  • Posted by jason (92.157.64.x) on April 24 2026, 5:39am in 8110:
    Moderators for the stash?

  • Posted by Justin on April 21 2026, 9:05pm in 8102:
    added GetUserFileName() in the latest +dev, can ask for open/save/dir/multiple open

  • Posted by Justin on April 20 2026, 8:09pm in 8107:
    it's the googles and the facebooks that will probably rise from the ashes, since they have the users :/

  • Posted by Mespotine (212.122.61.x) on April 20 2026, 10:08am in 8107:
    I think the user data will be spammed and spoiled by ai-input as well, so it's value will probably become less and less, once people realise, the data is full of ai-slop. But we will see, whether ai will survive in the first place, since it's not sustainable (yet) as a business model in the first place. So maybe most of these companies will not exist long enough to be their own pallbearer...

  • Posted by Dens (126.224.181.x) on April 17 2026, 9:52am in 8095:
    oulIf I may, there is a package called Reaper Update Utility on ReaPack that works well, but a native feature wouldn't hurt, especially if it skips the Installation window process (if such thing is technically possible)

  • Posted by Justin on April 15 2026, 4:47pm in 8104:
    non-jewish people sometimes choose to become jewish. again the ambiguity of language (the lack of distinction between ancestry and religion) is a problem here

  • Posted by ilovethem (154.161.121.x) on April 14 2026, 7:03pm in 8104:
    If your ancestors were not Jewish, then you wouldn't be Jewish. Is that not so?

  • Posted by ilovethem (154.161.121.x) on April 14 2026, 6:54pm in 8104:
    I like conversations like this as it makes me see things from others perspective. But your statement "we all are who we are, not who are ancestors are" as nice as it sounds is strange, cuz if ur ancestors are black then it's part of who you are now. So it becomes, "we are who we are because of our ancestry, and we can't change or reject it, we can only embrace it."

  • Posted by Justin on April 14 2026, 1:36pm in 8104:
    “sad”? no. proud? it is what it is, not sure where pride comes into play. we all are who we are, not who are ancestors are. “are you jewish” is an ambiguous question; it can be a question about religion or one about genetics, so I was answering both. as far as culture goes, that’s probably more complex, since it becomes a question of how much exposure I’ve had, etc.

  • Posted by Theyarenotperfectbutilovethem (154.161.60.x) on April 14 2026, 2:35am in 8104:
    And your answer above seems to imply that you don't separate the word "Jewish" from their religious culture of Jews as you answered by saying you are an atheist when the question is about you being Jewish?? I'm just trying to understand your logic about your answer???

  • Posted by Theyarenotperfectbutilovethem (154.161.60.x) on April 14 2026, 2:30am in 8104:
    Are you sad about your Jewish root or you are proud of it? Not trying to be political or racial but sometimes your comments sound like you distant yourself from Jewish religious culture stuff??

  • Posted by Mespotine (185.238.219.x) on April 13 2026, 3:31pm in 8102:
    +1

  • Posted by Wyatt (66.171.94.x) on April 11 2026, 2:25am in 8102:
    This would be great if you could add that to use in a lua script. Much appreciated! Thanks!

  • Posted by Justin on April 8 2026, 5:05pm in 8098:
    ah yeah, I guess it does. but funny I didn't read it that way, I think of it more like "real estate agent" which was probably the intent. but the multiple meanings is indeed funny!

  • Posted by Sunblister (77.187.168.x) on April 8 2026, 5:02pm in 8098:
    In many languages, "agent" has the connotation of "spy". The more neutral term would be "acteur" or derivatives. That's why I think it is funny

  • Posted by Beck (50.80.188.x) on April 5 2026, 2:53pm in 8094:
    The company gives the colective more rights as a legal entitty than any of the individial alrtists alone. Kind of like a frame to fit a cutural box assuring that we havet he "capital " and legal agency to actually get people away from "trading their hours for a hand full of dimes." In the name of true sovreighnty for the indiviuals. Not another go at the capitalism/ consumption lifestyle run. Having been extorted enough to know corperations have more life libertyPOHThanAnyIndividualSinceLifeIsBasedOnCapital

  • Posted by ikhider (169.150.252.x) on March 23 2026, 10:30am in 7967:
    Butch Ware for Gov! butchware4gov.com/

  • Posted by nicot (221.215.214.x) on March 22 2026, 6:19am in 8087:
    Yes, this method does work. But there's one issue. When multi-channel audio, such as 5.1 surround sound, is embedded in a video file, my hardware doesn't support multi-channel playback. Standard consumer audio devices can only output stereo sound. So is there a way to downmix all the channels into stereo? Well-known mainstream media players can do this natively in regular use.

  • Posted by Justin on March 22 2026, 2:54am in 8086:
    not all ffmpeg options are supported, yet

  • Posted by nico (221.215.214.x) on March 22 2026, 2:38am in 8086:
    If these parameters can be used with such flexibility, can I render container formats that are not natively supported in REAPER via these parameters? For example, .aac, .ts and other similar formats, just like this: format=adts codec=aac b=320k profile=aac_low ar=44100 ac=2

  • Posted by wasereb4 (95.91.78.x) on March 21 2026, 5:51am in 8086:
    nice!

  • Posted by Justin on March 20 2026, 9:05pm in 8086:
    codec=vvc or codec=libvvenc

  • Posted by Justin on March 20 2026, 9:04pm in 8086:
    then decoding should just work, and you can enable it encoding by using codec=h266 or whatever the encoder name actually is

  • Posted by nicot (221.215.214.x) on March 20 2026, 5:16pm in 8086:
    Yes. FFmpeg fully supports it.

  • Posted by Wah pedal (102.176.65.x) on March 19 2026, 10:33pm in 8084:
    I meant the tracks you produced and hosting on this site

  • Posted by Wah pedal (102.176.65.x) on March 19 2026, 2:59pm in 8084:
    Yeah, it's about time...lol. Love the sound of them. Is there a reggae track in your music collection?

  • Posted by Pow() (197.251.240.x) on March 17 2026, 10:14am in 8082:
    Alright

  • Posted by Justin on March 15 2026, 3:47pm in 8081:
    ah right on!

  • Posted by bot___ (212.69.51.x) on March 15 2026, 11:33am in 8081:
    Got it working with a Lua helper script to track the mouse movement over the video preview window. forum.cockos.com/showthread.php?t=307612

  • Posted by Sunblister (176.3.49.x) on March 8 2026, 3:08am in 8072:
    Yeah it 's good to see you guy skeep your humor amidst all the sad ness. My compliments to Ollie

  • Posted by Himi (49.43.90.x) on March 4 2026, 9:11am in 7966:
    here's a prototype I made back in Jan, what do you think? d1uia071p9z93f.cloudfront.net

  • Posted by JSFX ternary (102.176.65.x) on March 3 2026, 8:03am in 8067:
    Okay. Thanks.

  • Posted by Justin on February 26 2026, 5:48pm in 8063:
    ah, that's just x^4

  • Posted by NAD(name) (147.235.222.x) on February 26 2026, 12:13pm in 8063:
    I'm quite certain you're referring to the "Default" fader shape rather than "Max precision at +0 dB (4.0)". Just noticed btw that the faders in the native Cockos plugins obay the settings chosen in the TCP preferences page, that's neat. Don't get me wrong, x^3+x^6 is a lovely taper but it's nowhere as unique as the one I'm talking about.

  • Posted by Justin on February 25 2026, 1:16am in 8051:
    (though to get it to work properly on macOS, we had to get our audio worker threads marked as realtime in the correct way, which was a bit of a fuss!)

  • Posted by Justin on February 25 2026, 1:15am in 8051:
    That's not entirely true, if the OS is aware of the e-core vs p-core, and the DAW sets the priority level of its threads, then the scheduler can in theory do the optimal thing. I'm not sure of how well it works on Windows, but on macOS the OS does a great job.

  • Posted by Justin on February 23 2026, 3:36pm in 8059:
    I'd up-convert to 32-bit float, and then hope that the audio system does the appropriate level of dither on the output side

  • Posted by fladd (84.87.223.x) on February 23 2026, 10:33am in 8059:
    I understand that this is the case on windows, but I'm also targeting Mac (OpenAL) and Linux (PulseAudio, OpenAL). I think the situation there is better.

  • Posted by Justin on February 22 2026, 10:27pm in 8059:
    last I checked, DirectSound goes through the windows mixer, which does some really horrible things to the audio, so I think any effort you do will be lost, no point in dithering if the Windows mixer will resample your audio poorly after the fact.

  • Posted by fladd (84.87.223.x) on February 22 2026, 6:43pm in 8059:
    Another thing I am considering is to convert to 32-bit float myself before giving the data to the audio system. This is mainly because not all of the audio systems I need to target support all formats natively. In particular, OpenAL does not support 24/32-bit integer). So my options here are to either down-convert to 16-bit (potentially with dither), or to up-convert to 32-bit float. Not sure what makes more sense here...

  • Posted by fladd (84.87.223.x) on February 22 2026, 6:41pm in 8059:
    Thanks for the response! I am using pyglet for the implementation and hence have only access to DirectSound and Xaudio2 on Windows, OpenAL on Macos as well as OpenAL and PulseAudio on Linux. I also don't have an influence on the final conversion from each of their internal 32-bit float representations to the DAC output bitdepth. I could only dither on the way into those systems (i.e. before they get converted to 32-bit float; for instance by converting a native 24-bit stream in to a 16-bit stream).

  • Posted by PThread (83.30.240.x) on February 21 2026, 8:28am in 8051:
    True - however, for now Reaper can't fully utilize all cores, because P‑cores and E‑cores have different capabilities. It's not just an Reaper problem; awareness of hybrid CPUs is still limited.




[unreplied] | [replied] | [recent comments] | [all]
[older questions]
Copyright 2026 Justin Frankel. | RSS