Hi
I’m working on a live set and have two questions:
Do you guys have an entire set worked out in one project (instant switch, might be cramped with sample time) or one project per “song”? If in separate projects, do you find the loading time problematic?
If I start with preparing each “song” in separate projects, is it possible to merge (sample data and sequence data) into a single project later on?
Definitely put your whole set in only one project!
Loading a new project live isn’t a very good idea. It takes time and you may have trouble starting it in sync with whatever you have playing to fill the (silent) gap inbetween loading projects on the DT.
Also: Where is your master-clock coming from?
Merging projects is impossible afaik (I do not use overbridge, so might be wrong on that part), but you may use elkherd to reorganize patterns and samples in a single project.
I highly recommend organizing in a one song per bank fashion. Although some people seem to be able to even have one song per pattern…
I described another way to prepare transitions of patterns here: Planning a Syntakt live set - transition strategies - #7 by Pablo76
Written for the ST, but applies for the DT as well.
I’m interested in this as well. I busted out DT/DN combo and I’m trying to figure out how I should go about this.
With song mode you can create different songs per project and load them without stopping the current song. Isn’t it?
If possible put it all in one project.
You can’t transfer patterns from one project to another with Elektron Software unfortunately. ElkHerd is a third party app for that but only supports DT 1.4 (without being able to transfer song data) atm. It’s written by @mzero in his free time.
You can copy patterns across projects on DT itself but it only copies over sample slot numbers (without the actual samples which might differ from project to project).
hopefully he doesn’t mind me posting this but I think @Doug made this all as one project with DT and novation circuit rhythm and it def plays like a set (to my ears at least), maybe he can comment on a strategy for you.
I play all in one pattern. Everything improvised. If i would preplan different songs i would use one project.
With 1.50’s RLen feature on the record page and with nimble fingers (especially tapping along to the beat next to the button in advance of the 1.1 beat), it’s now technically possible to live-resample a playing pattern in mono, drop it into a pre-configured track, filter or loop as desired, and leave it playing through pattern change (by muting the corresponding track on the next pattern). Not Octatrack-caliber stuff, but there are definitely some interesting DJ-style possibilities there…
Yeah, that one is all one long pattern chain in one project. Most of the samples are on the Circuit Rhythm for that one, which has a lot less sample time per project. I think it’s like 220 seconds.
I’ve never maxed out a project in terms of sample time. I have run out of sample slots and had to purge unused samples. However, with the 1.5 update it’s a lot easier to set up things like sample chains. So even that is less of an issue.
Long story short. No matter what kind of set you want to build, you can and almost certainly should build it on one project. And merging your patterns later is doable but it can be kind of a pain especially if you use a lot of sample locks.
Thanks for all the answers. It seems there’s pretty much agreement on having everything in one project. I’ll proceed in that direction with confidence!
Thanks again!
I use one pattern per “song”, but more as a starting point that leaves quite a big room for improvisation (and error lol). For smooth transitions i will always copy some elements of the last track to the next one (For example a Snare(2), a Hat(6) and a Bassline(9) and use Pattern mutes so when i change the pattern you won’t hear a new element until i decide to unmute it. In the pattern before i write into the title wich tracks i copied to the next pattern, so i can bring those in before a pattern change. (in this case the pattern before would be named BANGER 269). So i guess you could say transitions are quite pre-planned, but everything in between is not. Works pretty good for me!
…usually a song per pank…gives u 8 songs to perform…that’s at least half an hour, more likely 40 minutes, of fluent stage time…
no floor will judge u, if u break the flow for 20 seconds max, in case u have to upload a new project…
if u wanna play content from various projects, have little inbetween snippets prepared, coming from another source or catch the ending moment of a perfamnce segment in some kind of external reverbt/delay fx unit, shattering along while u get hands free to change the project container and make that switch…
but basically, never be afraid of silence…it’s powerful, if u use it to ur advantage in all sorts of sonic story telling and jsut make sure ur change project procedure is rehearsed and therefor as quick as possible…
wait, what?
I need to actually explore song mode (no time, and trying to write a whole live set… guess I’ll get there when I get to the “making songs” rather than “making patterns”!)
Yes, this works well, though the next song gets queued up for the next end-of-row, rather than playing to end of song. So it takes a little speed, or looping the last row of the current song.
Having put together some scratch arrangements of 2-3-pattern ideas, I’m averaging about 20 rows a 4-minute song, for building a little set for DT/ST in song mode, so I think I’ll be going song by song (or sometimes combining 2 or 3 and their transitions) to build up to a longer set.
This is already an old topic but I can give my take on this as this is pretty current thing I’m working on at the moment.
I have a melodic techno set on Digitakt which is currently able to play 2+ hours without streching too much. It is arranged in one project using pattern and banks quite liberally. If I remember correctly it spans around middle of bank F currently.
I don’t have that many long samples few pads and risers. Otherwise few kicks, hihat, clicks and clacks for snare, single cycle synth samples which many of them are re-used around the set. By tweaking sounds I’m able to get them sound pretty different and after all leads and bass lines many times in techno and trance utilise pretty similar sounding synths. That being said I haven’t run in any issues regarding memory use. I don’t know exactly how much is being used as I don’t monitor it because that really is not concern.
Structure of the set is to save pattern mute states in the patterns. When pattern is changed I know that they will go well together as state of mutes are set so that it flows nicely forward. It can’t be really thought as a songs as they merge seamlessly together. Some tracks go away and new ones are introduced along the way. Thematically I could say that main components usually are kepts couple patterns after they got replaced main lead, bass synth or some prevalent rhytmical structure.
Only thing that I’m still struckling is to build those jaw clenching long and massive risers. They are not necessarily needed in techno but if leaning more toward the trance I would like them. At least they work well on a dance floor. I have an idea to dedicate whole pattern for such long riser. Then slow down certain tracks to run 2x or 4x longer to accomodate riser. Currently I have just some small drops which are done using fill mode. They are pretty easy to place. If going to this riser per pattern approach I need to be more careful that I would not over shoot that climax and cue next pattern so that the drop is timed correctly. In regard to this longer risers and drops with Digitakt I’m open to new ideas if someone has some interesting techniques for that.
I think the idea of the separate pattern is a good one. You could also name it appropriately so you know what’s going on when you get to it.
I would trigger the riser with a “1st” condition, so if you accidentally forget to queue a new pattern the riser won’t repeat (unless you’d prefer it did?).
I would then personally use Per Track scale and set the Ch Len to however many steps you want the riser to end at. You wouldn’t need to actually set any of the individual tracks to different lengths, you’re just making use of the Ch Len feature from that mode, so that a queued pattern change happens after a specified number of steps. That way as soon as you move to the riser pattern, you can then immediately (or soon) queue up the following pattern and it will definitely happen after the time you’ve set.
Example - you’ve got a 16 step pattern, but want the riser to run over 8 cycles of that, i.e. 128 steps. Set the Ch Len to 128, condition the riser trig with “1st” so it only happens once, and your pattern will then run 8 times through as the riser builds, and then move onto the next pattern that you’ve queued.
At which point you’ve got 8 cycles to ad-lib any fun knob twisting during the riser, knowing that your pattern change is all taken care of!
That may be what you were already suggesting, but rather than “slow down” any tracks, this approach lets you just take a straight up copy of the main pattern (e.g. the 16 step one) and not have to mess with any of the steps in that to make it into a 128 step version with the riser and pattern change taken care of.
Thanks. Of course that chain length setting woul do the trick. I didn’t first thought that at all. There is no chance that transition would be missed. I can just hit pattern cue right after riser pattern starts and focus on knob fiddling and be certain that drop will happen at the right moment.
About naming, I use this rather cryptic naming convention for my patterns where I can indicate things that I should do to that pattern while playing e.g. F5U3LFO24. This would be Rise track 5 filter up, rise (unmute) track 3 and fiddle with LFO2 of track 4 to create some interesting movement. I try to avoid those mute/unmute directives as I already save mute states in patterns but sometimes like in previous example I would like to steadily open filter for pad and after drop in hihat from track 3 and continue evolving that pattern. Of course this could be saved to separate patterns but this is kind of balancing act between keeping performance simple and sparing patterns for future use.