Speeding up a song. Highest quality solution/algorithm?

[~BURNY~]

So I have a death metal track out of a bunch of others I'm working on.

So far only the drums have been recorded and I'm currently editing them.

I know the song needs to be pushed a bit. (175bpm to say 180~ish bpm)

Wondering if any of you faced this problem and what was the best/less destructive process.

Should I speed the drums inside the mix (using a stem obviously to not mess the phase) or use the full mixdown? What is the highest quality time-stretching algo available for this task today?

Or should I quantize the drums to a new tempo (not sure how to do that)?

 

[tedtan]

I haven't tested any time stretching software lately, so I don;t know what the best currently is. But if it were me, I would copy the drum tracks to a new project (so I still have the originals in case this doens't work) and stretch/compress it using the DAW's internal time stretching algorithm. Then I would listen back and, if I couldn't hear a difference from the original aside from tempo, I would call it good. If I could hear a difference in tone, I would scrap that and, using the copied drums, create a REX2 file using Recycle. That way, you would have the individual drum hits as "slices" that would be moved closer together without affecting the tone.

 

[wishtheend]

Export raw drums, and stretch all by equal amount prior to slip edit/quantizing. I'd recommend Izotope RX, but don't really go beyond 5-10bpm for the cleanest sound. I think trying to quantize without stretching would be a hot mess especially in the overheads, so stretch raw drums all equally and then start your edits from there.

 

[~BURNY~]

Export raw drums, and stretch all by equal amount prior to slip edit/quantizing. I'd recommend Izotope RX, but don't really go beyond 5-10bpm for the cleanest sound. I think trying to quantize without stretching would be a hot mess especially in the overheads, so stretch raw drums all equally and then start your edits from there.

I don't think this will preserve the phase relationship between the tracks, will it?

 

[MrBongo]

when i stretch separate tracks in Reaper, they won´t work together at all. Be it drums or stereo guitar (even from the same take). Seems like the stretch algorithm adapts to the source material, to achieve the best possible result.

 

[Nikolas Quemtri]

I think there's a pretty quick way to do that. You may quantize all the drum tracks at the original tempo, without any stretching, using just slip editing (dynamic split + quantize, actually), then change the tempo to a higher value (don't forget to set the "beats, length, rate" timebase in the project settings first). All the slices will have to be preserved on their initial places, but stretched now. Select them all (all the drum items), go to the item properties, and set the rate back to 1.0. And yeah, do not allow any padding space before the actual transient, while dynamically splitting the tracks, 'cause it may result in the biased transients after increasing the tempo.
Since you're speeding up the song, there should be no artifacts or doubling hits this way.

 

[~BURNY~]

I think there's a pretty quick way to do that. You may quantize all the drum tracks at the original tempo, without any stretching, using just slip editing (dynamic split + quantize, actually), then change the tempo to a higher value (don't forget to set the "beats, length, rate" timebase in the project settings first). All the slices will have to be preserved on their initial places, but stretched now. Select them all (all the drum items), go to the item properties, and set the rate back to 1.0. And yeah, do not allow any padding space before the actual transient, while dynamically splitting the tracks, 'cause it may result in the biased transients after increasing the tempo.
Since you're speeding up the song, there should be no artifacts or doubling hits this way.

I have no idea what DAW you are talking about
I'm using Cubase 8.5.

 

[Nikolas Quemtri]

I have no idea what DAW you are talking about
I'm using Cubase 8.5.

Sorry, dude, I thought, you're using Reaper too. Just mixed up your post with another guys'.
Anyway, the basic idea I'm trying to suggest is to quantize all the drum hits as separate slices/items, so they'd be shortened after the tempo increasing.

 

[KillFrenzy]

The phase correlation can be a problem if you mess with time compression in a multitrack scenario. Reaper has lots of algorithms and probably one of them could work, but since you're using Cubase I don't know what to suggest.
If you want me to try that for you, just send me the tracks and I can do some tests using different approaches and algorithms and send the track back to you.

 

[Machinated]

for drums, if you have the patience, I'd cut at each hit and push them closer together, then crossfade. I've done that for a similar tempo change for what you're trying and it's pretty transparent. Drum transients really can get squishy and weird with stretching.....

For the rest, people seem to suggest izotope and IRCAM have the best algo's.

 

[KillFrenzy]

for drums, if you have the patience, I'd cut at each hit and push them closer together, then crossfade. I've done that for a similar tempo change for what you're trying and it's pretty transparent. Drum transients really can get squishy and weird with stretching.....

I've done that and it didn't work very well, probably because of the rest of the elements of the song. It was a very clean mix. Some cymbals sounded a little choked. But it's a very good suggestion anyway.

Try doing a second cut after the drum hits and stretch only from that point. This way, the transients are completely preserved. Don't forget to do a small crossfade between the unstretched and the stretched part.

 

[mva801]

The way I would try it, in Pro Tools you can set a track to a tick timebase (tempo dependent) instead of a sample timebase. You could slice and quantize each hit, but don't crossfade yet. As long as the track is in tick timebase and each hit is on the grid, you can then change the tempo and each hit will stay with the grid. After you change the tempo, you can crossfade the tracks and consolidate.

Not sure how that would work in Cubase, but even in pro tools it would be a huge pain in the ass.

If time stretch algorithms didn't provide a usable result I'd probably have the band re-track it. I've done that more than once when a song just plain wasn't tracked well enough or at the right tempo. There are few substitutes for just getting it right at the source.

 

[~BURNY~]

Seems feasible but awfully tedious.

I think I'm gonna try to get Izotope RX and try it on a drum stem. Cubase HQ algos have been significantly improved but I'm not entirely sold on these.

Re-booking studio sessions is not an option for the band at this point unfortunately.

Thanks for the suggestions guys. Keep them coming.

 

[MrBongo]

Can you possibly share a demo of the song? Maybe there´s another way to make it work at 175bpm with a different arrangement

 

[~BURNY~]

Can you possibly share a demo of the song? Maybe there´s another way to make it work at 175bpm with a different arrangement

I would need the band's agreement to do so. I'll ask them.


Now believe or not, I just tried to raise the tempo in cubase and somehow it seems to work (all the drums have been previously quantized so everything is sliced and crossfaded). I need to investigate this.

 

[bryan_kilco]

Not too sure on how it would affect phase, but I was going to suggest simply bumping the rate up while preserving original pitch.

 

[Korwent]

Whatever you end up doing, do it on the full stem, or you will definitely mess up the phase!

 

[~BURNY~]

So yes, after all it seems that when every hits are sliced, quantized and crossfaded, I just have to increase the tempo and all the audio segments follow the change and it sounds pretty convincing. I haven't messed with the timebase and grid settings, it just happened to be properly setup right off the bat so results may vary. Also I always cut the hits with a slight margin ahead of the hits no matter if I have to move forward or backward, so the transients never get eaten by the crossfades.

 

[Korwent]

Yeah after thinking about it a little the phase issues you could face would be at a far higher frequency than what's audible (think Nyquist Shannon theorem), so you should be fine, except if your song is extremely short.

 

[~BURNY~]

Yeah after thinking about it a little the phase issues you could face would be at a far higher frequency than what's audible (think Nyquist Shannon theorem), so you should be fine, except if your song is extremely short.

Je me permets de te répondre en français:
Je ne suis pas sûr de te suivre là. Des différences de phase, il y en a par la force des choses entre les différentes pistes et elles affectent tout le spectre et donc sa partie audible quelque soit la valeur Δt. La difficulté est de préserver Δt tout au long du morceau et c'est apparemment impossible avec du time-stretching qui va créer des fluctuations (effet phaser).
Dans mon dernier post, le problème ne se pose plus puisque tous les segments bougent ensemble (d'une piste à l'autre s'entend) et donc Δt est préservé.