Why do Render and Resample give different results?

[b0unce]

its like I've said all along, these daws may all be dealing with the same bits and bytes and the same math rules....but you're still at the mercy of the coders, and how they handle the bits bytes and algorhithms etc...

the only point I wished to make is that its not IMPOSSIBLE that one daw could sound different to another.....whether they do differ or not, I dont know. but occasionaly you get guys saying things like 'hey its all maths, therefore they must sound identical' ...in an ideal world where bugs dont exist maybe its OK to make such absolute statements

[tomperson]

It's not just a matter of bugs, but it is also a question of design decisions. For example, when coding a "traditional" daw, the design team can just leave timestretching/pitchshifting as an offline process, hence, the development team can afford implementing an algorithm that eats as much CPU as needed. In live, this is different, the program must be able of doing that in real time, so there are quite different requirements involved. Trade offs must be made.

[Robert Henke]

I did not say there IS a bug, i said it is POSSIBLE* that there are very special conditions involving VSTi,VST, Audio units, ReWire, etc. etc. where things could go wrong. Inside Live, inside the other applications, or due to driver or OS issues.

We made tons of test, we take every sound quality discussion serious and if someone comes up with a reproduceable problem we will fix it asap as we did in the past.

Robert



*every system of a certain complexity can have unkonwn bugs. This has been proven by some very clever mathematics guy some decades ago.

[knotkranky]

Resample is the more accurate. Real time always is. There has to be some sort of interpolation in render. Time based effects must have a hell of a time rendering properly, compressor timings, reverbs, midi timing etc. I simply don't use it.

Render is really just another time compression algorithm stacked on the others. Not good.

I don't know why anyone would work on their project for days, weeks or months for that matter, only to do a non monitored blind high speed render to save a few minutes at the most crucial moment of printing a mix. Garageband does high speed rendering. ProTools does real time rendering. Hmmm.

I read so many of your posts that spread shite information. Rendering in non real time is nescessary for most native users. Im sure you can work out why. Offline bouncing has been around for years now and while there may have been a problem a few years ago on a couple of daws none have the problem today.
Being a pro tools user (you mention it constantly) is lovely and dandy but doesnt mean your an expert. I was once the proud owner of an HD3 rig complete with control 24. When the new accel dsp cards came out less than nine months after I bought my rig I was sickened to my sthomach to see my process cards being sold for less than 700 quid which paid three times that. I vowed then that would be my last with digidesign and six months later I sol my rig. Interesting to note my mixes turn out much better on native and thats ALWAYS with offline bouncing.

Render is really just another time compression algorithm stacked on the others. Not good.

That comment you made is utter utter utter shite. NO ONE LISTEN TO THAT YEhaa.

Time based effects must have a hell of a time rendering properly, compressor timings, reverbs, midi timing etc. I simply don't use it.

You even admit to never using it so how do you know its shite. Nearly everything I do is midi and timing is crucial. I would have thought that after many years working I would know if rendering had affected midi timing. Your just plain wrong and you need to stop spreading shite.

You trip on the things you don't like about people and rarely focus on the points that are trying to be made in the positive. Throw away the negative bias and my message will be clearer.

Yes, I know why render is necessary. It sounds different to me than realtime is all. It is very subtle, but it is there. Will you sell less hits because of it? No

Yes, digidesign are bastards but they make a sweet daw. In my world, no PT, no business. I'm glad you found live.

Of course render uses time compression. When 5 minutes takes 1, it's time compression. Is it adding/taking audio away like warp? No. I can see how i went south on that. My bad.

I've used render a few times and don't like it. Understand I am hypothesizing and looking to be educated on this topic. If I'm spreading shite, i'm only spreading shite questions and hypothesize, which is only true if you believe so.

Prove that I can hear the difference? No. I'm not here to be right. I'm only here to get some answers to my musings. Render works nicely for you. Great. It appears we are both getting what we need from live.

[Spiralgroove]

Resample is the more accurate. Real time always is. There has to be some sort of interpolation in render. Time based effects must have a hell of a time rendering properly, compressor timings, reverbs, midi timing etc. I simply don't use it.

Render is really just another time compression algorithm stacked on the others. Not good.

I don't know why anyone would work on their project for days, weeks or months for that matter, only to do a non monitored blind high speed render to save a few minutes at the most crucial moment of printing a mix. Garageband does high speed rendering. ProTools does real time rendering. Hmmm.

I read so many of your posts that spread shite information. Rendering in non real time is nescessary for most native users. Im sure you can work out why. Offline bouncing has been around for years now and while there may have been a problem a few years ago on a couple of daws none have the problem today.
Being a pro tools user (you mention it constantly) is lovely and dandy but doesnt mean your an expert. I was once the proud owner of an HD3 rig complete with control 24. When the new accel dsp cards came out less than nine months after I bought my rig I was sickened to my sthomach to see my process cards being sold for less than 700 quid which paid three times that. I vowed then that would be my last with digidesign and six months later I sol my rig. Interesting to note my mixes turn out much better on native and thats ALWAYS with offline bouncing.

Render is really just another time compression algorithm stacked on the others. Not good.

That comment you made is utter utter utter shite. NO ONE LISTEN TO THAT YEhaa.

Time based effects must have a hell of a time rendering properly, compressor timings, reverbs, midi timing etc. I simply don't use it.

You even admit to never using it so how do you know its shite. Nearly everything I do is midi and timing is crucial. I would have thought that after many years working I would know if rendering had affected midi timing. Your just plain wrong and you need to stop spreading shite.

You trip on the things you don't like about people and rarely focus on the points that are trying to be made in the positive. Throw away the negative bias and my message will be clearer.

Yes, I know why render is necessary. It sounds different to me than realtime is all. It is very subtle, but it is there. Will you sell less hits because of it? No

Yes, digidesign are bastards but they make a sweet daw. In my world, no PT, no business. I'm glad you found live.

Of course render uses time compression. When 5 minutes takes 1, it's time compression. Is it adding/taking audio away like warp? No. I can see how i went south on that. My bad.

I've used render a few times and don't like it. Understand I am hypothesizing and looking to be educated on this topic. If I'm spreading shite, i'm only spreading shite questions and hypothesize, which is only true if you believe so.

Prove that I can hear the difference? No. I'm not here to be right. I'm only here to get some answers to my musings. Render works nicely for you. Great. It appears we are both getting what we need from live.

you dolt, did you even read what mr. henke responded?
youre still arguing points that have already been refuted in this thread by everyone but you

[tomperson]

Of course render uses time compression. When 5 minutes takes 1, it's time compression. Is it adding/taking audio away like warp? No. I can see how i went south on that. My bad.

I find that misleading. There's not compression of anything. It's an offline process, you're comparing apples to oranges here. There's no concept of time on a batch/offline processing such as this, at least not in a sense that can be compared to the lenght of the piece if you played it in realtime.

If you had, say, a song with 100 channels, each with it's own group of VSTs and such, and that song was a 5 second piece, i bet that the render would take a whole more than 5 seconds to complete, hence throwing your hipotesis of time compression out the window...

(and, by the way, that piece would be certainly possible to be rendered offline, but probably impossible to play in real time)

[knotkranky]

knotkranky, please take note of what Robert said, unfortunately your former posts are full of misconceptions, you should not say things as if they were facts when in reality, YOU DON'T KNOW FOR SURE. That only helps spreading misconceptions.

Take it as an advice, not as a personal attack.

I've only stated facts about what I'm hearing. And yes, "I don't know for sure" about why, which is why I made invitations to be educated. If only facts were allowed in this forum, where would all of us be? Mr. Henke's thoughtful post is helpful to me but it still remains that rendering "can" sound different from realtime. Even realtime mixes can have differences depending on how one has set lfo's, right? Circumstances can make a mix unique from render to render involving third party plugs, rewire, driver issues bla bla bla. And yes, this is true for all native apps and is not a live only issue. This is why PT HD is closed, proprietary and freakin expensive! Regardless of all that, my ears tell me what to use.

Much of my style has to do with the fact i've been doing this for a long time. Digital problems annoy me a lot. They can be so subtle like degrees of jitter and such. That kind of listening is no fun at all. If my dig mix copy is not exact, It's onto another way. I've got a couple heartbreak stories concerning these issues so now I'm a tweeky digital bastard.

[knotkranky]

Of course render uses time compression. When 5 minutes takes 1, it's time compression. Is it adding/taking audio away like warp? No. I can see how i went south on that. My bad.

I find that misleading. There's not compression of anything. It's an offline process, you're comparing apples to oranges here. There's no concept of time on a batch/offline processing such as this, at least not in a sense that can be compared to the lenght of the piece if you played it in realtime.

If you had, say, a song with 100 channels, each with it's own group of VSTs and such, and that song was a 5 second piece, i bet that the render would take a whole more than 5 seconds to complete, hence throwing your hipotesis of time compression out the window...

(and, by the way, that piece would be certainly possible to be rendered offline, but probably impossible to play in real time)

I've admitted my mistake in the very quote you posted.

[Hepha Luemp]

Hi all!

I was thinking of maybe introducing a new element into this discussion, spice it up a little maybe?

I thought I could share a memory with you, and yes! - I think I found one of the original threads I was thinking of.

Follow the link for a long discussion on whether a digital DAT clone sounds identical or not.

I not kidding!

My favourite is the posting that says that although they are identical number by number, they still sound different!

Well, I can imagine the possibilty, - why not?

Here's the link:

http://groups.google.com/group/rec.audi ... 784401360e


By the way, I also think that Live 6 sounds a little bit different from 5, a little bit smoother maybe?
Hehe!



Hepha Luemp

[Tarekith]

I've already tested and shown that renders in Live 5 and Live 6 are bit for bit identical. Just FYI.

[Tarekith]

Hepha -

Interesting read, but after an admittedly quick skim of an 11 year old thread ( ), it looks like the differences they are hearing are due to jitter, which is NOT an issue at all for what we aer talking about.

Actually it might possibly explain why the Kranky guy is hearing things no one else is, but that only means his DAC is very old and prone to jitter. In which case I find it ironic he does mixdowns for a living, so I'm not inclined to believe that's the issue.

[nolus]

I don't know why, but I feel obliged to sspeak up in knotkranky's defense. If he says he hears a difference between render and resample why doubt him.

I know of at least one vsti that can sound different when rendered off line, and i've no reason to think there aren't others.

I use the ivory piano vsti quite a bit. Because it is such a hungry beast it is sometimes necessary to limit the popyphony. But when rendering off line it always uses unlimited polyphony. as a result the rendered version can sound muddy and cluttered compared to the real time performance.

it is intended as a helpful feature, but the fact is if you have created a performance that sounds right to you in real time then you might well feel that the rendered version is unfaithful to the mix you created in real time. I would not be suprised if many vst plugins, especially "A list" ones, don't use similar tricks to provide "improved" quality when rendering!

[Johnisfaster]

not sure if anyones brought this up but hasn't anyone considered that things like reverbs and lfo's and tons of other tools that may be in your song use complex algorhythms that are not going to produce the same results every time you render, so render your song 10 times and none of those are going to cancel each other out most likely. doesn't matter if it's the same rendering process or not if you're using a chorus or flanger or something like that then you are likely going to get different results with every render.

if this has been mentioned then ignore what I said.
I was planning on reading all 4 pages but I got lost around the time people were debating symantics and "computers guessing" things.

[Tarekith]

I don't know why, but I feel obliged to sspeak up in knotkranky's defense. If he says he hears a difference between render and resample why doubt him.

I don't have issues with what he hears, I have issues with illogical explanations of why these are different, and what rendering is.

[Tarekith]

not sure if anyones brought this up but hasn't anyone considered that things like reverbs and lfo's and tons of other tools that may be in your song use complex algorhythms that are not going to produce the same results every time you render, so render your song 10 times and none of those are going to cancel each other out most likely.

if this has been mentioned then ignore what I said.

Yes it's been mentioned, my test specifically excluded those types of things for that very reason.