Scary thread about Live on KVR - please address!

[jlgrimes]

I give up.

• Sonar's midi timing does seem tighter than Live's mainly when working with external modules. My Studio Electronics SE1x always had some minor latency in Sonar, but in Live it seems even worse sometimes to the point I rarely record with it anymore. As kb420 said, I can actually see the latency easy in Live when tracking out the external midi.
  
  I notice some minor inaccuracies when midi recording softsynths as well, but it is pretty minimal at 2.9 ms of latency. It is esp a non issue if you auto quantize everything, but it would be nice to have your original timing just as you played it.
  
  With audio, I always record with monitor off and have no issues. With monitor on I use 2.9 ms of latency and have no issues.
  
  What I do notice is when tracking midi out at higher latencies, external midi modules will be very audibly off. I remember an old version of Pro Tools having this problem as well.
  
  I think this is more of an ASIO problem. On Sonar with WDM drivers I can record at any buffer setting and always get the same latency which makes recording more managable, but to be fair ASIO in Sonar works a little different in Live. Live's ASIO recording seems better for recording audio, but I only have problems when tracking out midi.

[gurumonkey]

I agree that this is one of the most fascinating threads i've read. learned a lot for sure.

the real question is about real time monitoring, so it seems.

and the question probably depends on what style of playing you're accustomed to.
let's use guitar rig as an example. don't have it, but that will work.

Some people, when monitoring through the software, aren't terribly concerned with the monitored sound lines up with the recorded sound, but they still want to monitor directly because they want to get a feel for the kind of effect they are using. Since monitoring is really only a mood gauge, they tend to ignore the exact timing of the sound they are hearing monitored and instead would focus on making sure their pick strikes the guitar string at the exact moment they hear the beat from whatever other tracks they are playing. We'll call these people group A.

Some people, when monitoring through the software, focus DIRECTLY on the timing of what they heard They don't direct monitor to pick up the mood of the plug in they are using, they want to make sure that the sounds they hear whilst they are playing directly are the sounds that recorded alongside the other tracks that they are hearing. Recording precisely what they hear is absolutely most important, and they are willing to adjust what time the pick strikes the guitar string in order to hear the finished product as they play. We'll call these people group B.

Both groups have been involved in this thread. I can understand both viewpoints. Group A is expressing unhappiness because LIVE does not currently meet their needs without recording two tracks, one while monitoring through the software and one with monitoring off. It may seem silly to some, but while i can simply listen to my electric guitar accousticlly, or though an interface's 0 latency monitoring solution, it doesn't quite feel the same playing a guitar solo with some sweet patch on it. But, this group of people would still prefer to make sure that their pick is what dictates their timing, not the sound heard. Certainly an option to change the compensation would be useful to such people.

Group B would be Ableton's take on the situation, as well as many musicians that focus more on sound than touch. They are already happy.

It's a matter of playing style.

[gurumonkey]

just to clarify, in case that didn't make sense,
the real issue is whether a person considers Touch or Sound to be more important. Touch would be group A, Sound would be group B.

the real danger lies when neither is most important and you end up in the middle. then you'd be frustrated no matter what the abes did.

[fishmonkey]

heheheheh, for a minute there i thought you were about to make everything crystal clear, but, well, no...

nice try, but the "touch versus sound" argument will surely breathe new life into this thread!

long live the thread!

[Tone Deft]

nice one gurumonkey. the details were TL;DR for me, I appreciate the summary.

is the core of the issue is the latency between routing through FX and not routing through FX?


FWIW I turn Live monitoring off, get all my guitar/whatever tone outside Live, work at a high sample rate and use the direct monitor from the sound card... and warp the shit out of all my takes anyway.


at the end of the day, if other DAWs aren't broken...
(show me a perfect piece of software and I'll show you a liar.)

[gurumonkey]

yea, i think that's really the issue here.

i'm with you. i have way too many expensive effects pedals to not use them at all.

[mdk]

nice summary. i was thinking how to express it better but you did the job nicely.

im definitely group A.

[deva]

What I have learned from this thread:

There is no bug...
Ableton made a choice in implementation that I now understand better...
Everyone can have it how they want and some people need to use 2 tracks to have it...
Nobody would have to use 2 tracks if there were 3 options instead of 2
1 Auto
2 Off with offset
and a new 3rd option
3 Auto with offset

[forge]

just to clarify, in case that didn't make sense,
the real issue is whether a person considers Touch or Sound to be more important. Touch would be group A, Sound would be group B.

the real danger lies when neither is most important and you end up in the middle. then you'd be frustrated no matter what the abes did.

I'm not sure about this really - I would say maybe you should have included a 3rd group: people who just do what they need to do to achieve the task at hand, which IMO should be everyone who makes music

In my experience things like latency, buffer size - any of the settings in the preferences in fact - are much more useful if they are seen as constantly changable

so for example, you need ultra low latency to play guitar, then temporarily have only the absolutely essential tracks and functions playing to save CPU so you can set the buffer size as low as possible to give you the lowest latency possible, record your guitar, then when you have recorded the part go back to how you need it for production

likewise, if you are trying to record an instrument and the current settings don't cut it, just change it, then change it back!! none of this really takes long - about as long as plugging in the lead and just part of the process

likewise, if you need to set up another MIDI track - that's no big deal either and will probably only be the case for a small part of the entire production process

I have been using Live since v1, and came to it after using most other DAWs for a long time too, - some of them when it was just MIDI and the computers were so slow that there were REAL latency problems, and I can't say I've ever found this situation to be a problem with Live

I just don't see it that you have to set it one way and leave it that way forever

[[nis]]

[nis] I have a question here, is jbone right about this at all? It seems like he's not, but if you could clarify that would be cool.

I wouldn't say he's wrong. He's got a point. He's only wrong when he says that Live adds latency. It doesn't. Your audio interface adds latency. You either have to deal with it when you monitor through software or you waive goodbye to using software-based audio effects and use a direct monitoring setup or a hardware mixer + hardware effects.

What jbone and rhythminmind are asking for is a function that auto-removes your soundcard's latency after the recording. I can understand that people are inquiring such a function as it might be handy sometimes, but you should be aware that what this function does is plain wrong, because the result sounds different afterwards on playback than during the recording/monitoring stage.

Let me pull up another example: Let's assume you are a DJ with 2 turntables and you are playing in a rather big venue, but the stage monitors are broken / non-existent and you have no headphones. We also assume that both records have the exact same tempo so you don't need to beatmatch them. Now the record on the left turntable is playing through the main PA speakers from the dancefloor. The DJ booth is 20 meters (roughly 60 feet) away from the dancefloor. This means that you hear the record with a latency of approx. 65 milliseconds. If you now listen to the sound at the DJ booth and you try to bring in the second record, you will notice that it is terribly delayed and your mix sounds wrong. This is because the sound of record A is your reference. As record B needs the same time to travel through the speakers and then back through the room to your DJ booth, it will be 65 ms off. If you would record this lousy performance on a tape, it would reflect what your audience and you heard: a bad DJ set (even though it wasn't your fault). If you want to deal with the drama, you would need to nudge record B ahead of time to compensate the latency. The mix will be perfectly in time then. Note that this is your one and only chance. There is no other way. If you would deliberately play all your records at the "wrong time" and want the tape recorder to automatically shift record B ahead of time, it would mean that the listener who listens to the tape afterwards would believe that your party and your mix was great, but it wasn't. Or the other way around: if you compensate the latency by nudging record B to match the timing of record A and your tape recorder adjusts the latency as well, then your party crowd would be happy, but your recording is wrong.

You are facing the excact same problem in a monitoring scenario in Live. The only way to get it right is either to compensate the latency yourself or use a direct monitoring setup.

Makes sense?

I get what you're saying, it's one reason why I think in order to use FX on my vocalist live, I'll have to get her to buy a couple pieces of hardware.

Vocals are tricky. Singing through your computer, which means hearing yourself late is confusing and indeed hard to compensate. I'd suggest to use hardware effects. Another solution might be a DSP based audio interface, such as a Metric Halo, Creamware/Sonic Core or one of the newer MOTU interfaces. They have very tiny roundtrip latencies for their DSP effects as you can apply these effects directly on the DSP mixer (before the sound actually goes into your computer).

Think about it like this: suppose you take a "real instrument", and mic it up. Lets say a piano. Then you record it with monitoring.

A real instrument will be monitored through the air. Recording it via a mic is possible (set monitor in Live to off), but applying effects / monitoring through the software inducts latency which you simply can't compensate.

[SWAN808]

I agree that this is one of the most fascinating threads i've read. learned a lot for sure.

the real question is about real time monitoring, so it seems.

and the question probably depends on what style of playing you're accustomed to.
let's use guitar rig as an example. don't have it, but that will work.

Some people, when monitoring through the software, aren't terribly concerned with the monitored sound lines up with the recorded sound, but they still want to monitor directly because they want to get a feel for the kind of effect they are using. Since monitoring is really only a mood gauge, they tend to ignore the exact timing of the sound they are hearing monitored and instead would focus on making sure their pick strikes the guitar string at the exact moment they hear the beat from whatever other tracks they are playing. We'll call these people group A.

Some people, when monitoring through the software, focus DIRECTLY on the timing of what they heard They don't direct monitor to pick up the mood of the plug in they are using, they want to make sure that the sounds they hear whilst they are playing directly are the sounds that recorded alongside the other tracks that they are hearing. Recording precisely what they hear is absolutely most important, and they are willing to adjust what time the pick strikes the guitar string in order to hear the finished product as they play. We'll call these people group B.

Both groups have been involved in this thread. I can understand both viewpoints. Group A is expressing unhappiness because LIVE does not currently meet their needs without recording two tracks, one while monitoring through the software and one with monitoring off. It may seem silly to some, but while i can simply listen to my electric guitar accousticlly, or though an interface's 0 latency monitoring solution, it doesn't quite feel the same playing a guitar solo with some sweet patch on it. But, this group of people would still prefer to make sure that their pick is what dictates their timing, not the sound heard. Certainly an option to change the compensation would be useful to such people.

Group B would be Ableton's take on the situation, as well as many musicians that focus more on sound than touch. They are already happy.

It's a matter of playing style.

This is a nice summary of the 2 ways of looking at this issue.

1. If the MIDI track's monitor switch is set to AUTO / IN, Live records the note exactly where you played it. This suggests that you need to manually compensate your play when you use a software instrument.

2. If the monitor switch is OFF, then Live records the note and shifts it ahead of time (by the amount of your audio output latency).
.

Something seems strange here - in the as far as I can see - in situation 1 Live DOES NOT record the note as you PLAY it but as you HEAR it through the system. IE it shifts the note you play BACK in time according to the system latency.

In situation 2 Nis says that with monitoring off - Live shifts the note FORWARD. However it seems to me the opposite. When you dont monitor - Live records what you play on the keyboard/drumpad. When you DO monitor - live records what you play - then shifts the MIDI back according to the total system latency.... which = what you heard. LIVE PREDICTS WHAT YOU HEARD BY SHIFTING NOTES FROM WHAT YOU PLAYED ACCORDING TO THE OVERALL SYSTEM LATENCY

I know he is an Ableton employee - but his explanation quoted above just seems wrong to me. Im not 100% certain that Im right - but all my simple tests have show to me that what I describe above is whats happening... perhaps its arrogant to suggest-but Im no fool and have been producing for over 10 years...its a bit worrying that there is this level of confusion over the issue even when explained by an Ableton employee....The fact that Nis had to give up this discussion I think lies in the above understanding I pointed out...which I believe is innaccurate...and this way of understanding things makes it impossible to discuss rhythminminds concerns - because Nis fundamentally is understanding the Live concept differently than rhythminmind. The argument is impossible to continue for them...

Live is the only one that requires manual offsetting after recording.

Yes I agree with this - there is a factored delay AFTER recording in line with total system latency.

There is still latency while monitoring thats with every DAW. Let me say this again monitoring only.

There you said it. Whilst you are monitoring you have latency. The audio signal you feed through a software will be late by the amount of the roundtrip latency latency of your audio interface. This means that what you play is not in time. Live does not add any latency, it records what you play. If you want to be in time with the previously recorded track, you need to compensate your play.

I understand the Ableton way of dealing with latency because it covers ALL system Latency that Logic and others do not....IE by recording whilst manually adjusting your playing - you can take into account the total system latency - whereas if you do not do this in Logic etc - there will ALWAYS be system latency (total DAW latency) added in Logic which you cant get around as said above...

There is clearly a different way of doing things in Live as opposed to other DAWs due to Lives 'Live' concept.

[[nis]]

1. If the MIDI track's monitor switch is set to AUTO / IN, Live records the note exactly where you played it. This suggests that you need to manually compensate your play when you use a software instrument.

2. If the monitor switch is OFF, then Live records the note and shifts it ahead of time (by the amount of your audio output latency).
.

Something seems strange here - in the as far as I can see - in situation 1 Live DOES NOT record the note as you PLAY it but as you HEAR it through the system.

No. You are talkng about the sound you hear from your plugin, not about the note.

[[nis]]

You are talkng about the sound you hear from your plugin, not about the note.

I'll create a little demo Live set later tonight or tomorrow to demonstrate this.

Or you could test this quickly yourself:

-create 4 tracks, track 1 contains a drum loop which you use as your timing reference (you could also use the metronome of course). Track #2 contains a MIDI clip with a note on every 8th or 16th. Track 3 and 4 will be your recording tracks, one with monitor on, the other one with monitor off.

-route the MIDI output of track #2 to a virtual MIDI port and set up track #3 and #4 to receive MIDI from the input of the virtual port.

- load an instrument with a percussive sound on track 3 and 4.

- now, if you playback the clip on track 2, it should play the percussive sound on track 3. It will be late, due to the delay of the outgoing MIDI note and the latency of the instrument device, so let's add a negative track delay on track 2 so that the percussive sound lines up perfectly with your drum loop when you playback the set. This should in theory be 2 times the output latency (or resp. the size of the overall latency), provided that your audio driver calculates the latency values correctly.

Ok, now we have a scenario where a keyboard player listens to the drum loop or the metronome and plays his MIDI notes as accurately as he can, which would be right on time with the drum loop.

Now record-enable track 3 and 4 and record the MIDI notes. You will notice that the track with monitor on will be right on time. The other one with monitor off is shifted by the amount of output latency. Test this, read my other post again carefully(!), then decide if this is right or wrong.

Best,
Nico

[fishmonkey]

i love this thread

i need a holiday

[SWAN808]

You are talkng about the sound you hear from your plugin, not about the note.

I'll create a little demo Live set later tonight or tomorrow to demonstrate this.

Or you could test this quickly yourself:

-create 4 tracks, track 1 contains a drum loop which you use as your timing reference (you could also use the metronome of course). Track #2 contains a MIDI clip with a note on every 8th or 16th. Track 3 and 4 will be your recording tracks, one with monitor on, the other one with monitor off.

-route the MIDI output of track #2 to a virtual MIDI port and set up track #3 and #4 to receive MIDI from the input of the virtual port.

- load an instrument with a percussive sound on track 3 and 4.

- now, if you playback the clip on track 2, it should play the percussive sound on track 3. It will be late, due to the delay of the outgoing MIDI note and the latency of the instrument device, so let's add a negative track delay on track 2 so that the percussive sound lines up perfectly with your drum loop when you playback the set. This should in theory be 2 times the output latency (or resp. the size of the overall latency), provided that your audio driver calculates the latency values correctly.

Ok, now we have a scenario where a keyboard player listens to the drum loop or the metronome and plays his MIDI notes as accurately as he can, which would be right on time with the drum loop.

Now record-enable track 3 and 4 and record the MIDI notes. You will notice that the track with monitor on will be right on time. The other one with monitor off is shifted by the amount of output latency. Test this, read my other post again carefully(!), then decide if this is right or wrong.

Best,
Nico

Hi Nico - I'd really appreciate if you could create this set for me in a Live 6 compatible format so I can see for myself. Better that way than us going back and forth trying to get it right...