ACE Studio #1: AI sulle mie tracce, nulla generato da zero
By SICK FOR SYNTHS
Summary
Topics Covered
- Using AI with parsimony and ethic, or nothing at all
- An assistant that loads tracks while you keep playing
- The AI has no ideas, only executes what you specify
- An arrangement in minutes: a layout, not a final mix
- It does everything, so even non-musicians can finish tracks
Full Transcript
Welcome YouTube friends, welcome to this new video.
I was contacted by the guys at ACE Studio and they suggested I try this software.
I work with artificial intelligence, but always, as I say, with parsimony and trying to do things with a certain ethic because I ask myself, if it does
everything, what do I do?
Sell everything and leave, basically.
But I was asked to try this DAW, and if I liked it, to review it, and guys, this in my opinion is a truly excellent way to use artificial intelligence.
So, artificial intelligence has been integrated into a DAW, which is ACE Studio.
But not only that: this can also interface with your favorite DAWs, so Logic, Cubase, Studio One, Ableton, whatever you normally use.
It integrates, so it manages to add tracks, do things, and work with all your complete instruments that you use simultaneously.
ACE Studio was basically born initially to create voices where you would input notes and words and it would sing by synthesizing the voice.
Instead, now a lot of things have been integrated, and so this will be a series of videos where I will first explain what it is and how it works.
In this case, I integrated it with ChatGPT because I want to use voice synthesis and talk to it instead of having to write.
This is certainly flashy for a YouTube video, but it also has its own utility.
Now let's take an old thing I had from a year and a half ago, an idea a friend gave me on a track I made for fun at his house.
I pulled out the tracks played by me inside Cubase, so they were played with the D-50, with the Juno-X, with the internal drums of Cubase, so I exported these tracks and let's see what we can do with them.
There's also his singing, a bit out of tune underneath, which in my head I can understand what he meant, which is very interesting, but listening to it like this is terrifying.
Let's see what we can get out of this thing.
So, let's see what we do today to leverage artificial intelligence in an ethical way.
Here we have ACE Studio now, and we are using an artificial intelligence agent which can be, as you see here, either ChatGPT or Claude or other agents.
Just copy this prompt below, paste it into our agent, and it will connect to ACE Studio.
You need the ChatGPT or Claude application installed on your computer; the web version won't work.
You need the ChatGPT or Claude application installed. In the case of
application installed. In the case of ChatGPT, which I am using now, go to Codex, okay? Not on classic ChatGPT, but
Codex, okay? Not on classic ChatGPT, but on Codex to activate your agent.
Once done, it will connect, and I'm using the voice chat in this case, okay ?
Which is down here at the bottom. Now
we're going to do something very, very cool.
It must have happened to you to have old projects, scattered ideas you hummed, or perhaps, like in my case, you gave a small backing track to a DJ friend of mine who can't sing at all, but sometimes he comes up with strange
things, melodies that I layer over, and maybe I get something out of it.
Maybe something gets sung over the top and it ends up being interesting, but developing work like that is complicated .
I can also sing something over it.
Obviously, in this case, one transforms it for you, etc., but we want to create it ourselves and have an AI assistant that works for us, not excessively, but gives us good help.
In this case, in Cubase, I saved some tracks of this project that I played myself and created myself, my friend hummed something over it, sent it back
to me, so we have a rough idea, okay?
It needs to be developed.
So, in the old method, I would have been writing the lyrics, the whole melody, deciphering the whole melody, doing the arrangement; in short, it would have been long and laborious, not that I don't know how to do it, but when it
comes to fairly standard music, there would be work to do, maybe an AI agent speeds it up for us, maybe those slightly mechanical, repetitive parts, which are very, very useful.
Now I'll show you what we can do just by setting up the voice. First of all, let's grab the audio. Listen to me, if you go to file open, you'll already be in a folder with some tracks to upload,
can you upload them for me, please?
Yes, I'll take a look and see what I find there. In Access Studio, I'll check
find there. In Access Studio, I'll check where you are and confirm which tracks I see before uploading them. Wait a second .
I could also upload only Access Studio, but not the folder that appears in file open. If you paste the path for me, I'll import them right away from there, otherwise tell me where the tracks are and I'll go get them.
If you see those tracks, import them all .
Okay, I'm checking them right now. Go
ahead, I'm verifying. I see seven files in the Access Studio import folder, I'm loading them onto separate tracks aligned at the start.
I'll let you know when the import finishes. Done, all seven imported onto
finishes. Done, all seven imported onto separate tracks aligned at the start.
Think about this thing I just did.
Now I could do it too, but maybe you're playing another synth and need to load a loop, setting it up without having to move because you're busy making sounds.
So you have an assistant who, in this case, has loaded a backing track on separate tracks for you.
For example, right now could I say, add an audio track for me and arm it for recording?
Yes, let me check and I'll let you know .
I've added the recording track, but to arm it, I need an audio input.
Which input do you want to use?
Input four.
Okay, I'm setting that up now.
You see, I can now just focus on playing , for example, right?
I've armed the recording track on input four, channel one.
It's ready to go.
Forgive me, digital four, okay?
Okay, correcting that now.
Remove the audio track for me.
Okay, removing it now.
Set, set, set the BPM to 115.
All right, I've set it. Set to 115 BPM.
Let's have a listen.
Okay, starting playback.
Starting from the beginning now.
I'll spare you the agony of my singing, but in my head, I can already tell it could be interesting, and now let's see the first thing we're going to do, adding some parts that might be interesting.
In this case, I can say this.
Add a track and load a synth for me.
All right, I'm checking in the studio now.
Because this also loads your favorites, including all your favorite effects, and it can mix the track for you.
Now I'll show you, starting from an idea you've already made and developed.
I'm inserting it on the new track and looking at the available string presets .
So, I'll open it up and show you what it does. If you go to Iconica Sketch, those
does. If you go to Iconica Sketch, those are excellent strings, guys.
Okay, I'm checking if I can find it in the presets.
Naturally, the execution speed depends a lot on your computer and your internet connection, and above all, the connection to their server, how fast it is at this moment, so it's normal that
it might take a little bit, but it's...
I've opened Halion Sonic. Iconica Sketch
is there, but the preset must be chosen from the internal browser. Inside there
you'll find Violin Sustain.
Should I set it, or do you prefer to choose it yourself from the list?
No, you set it.
Okay, I'll set it. Got it, now I'm loading it onto the string track.
I did it, I did it, you know. I did it, you know.
I've done it, let me check what settings remain.
Yes, I see Violin Iconica Sketch in Halion Sonic with staccato selected.
Let's step back a moment.
Okay, starting playback, starting the loop and re-sampling.
Put this backing track on loop for me.
All right, I'll set the loop.
Hit stop and let's start over.
Okay, stopping and restarting from the beginning with the loop.
Restarted from the beginning with a loop over the whole track, about 1 minute and 30 seconds.
I want to record a part, set it to record and let's go.
Okay, I'm preparing the recording and starting it up.
I've activated loop recording on the string track.
The backing track has started and is recording from the beginning. We're past
the count-in.
Okay, set a loop from 1 to 9.
All right, I'm setting the loop from 1 to 9.
Loop active from the start of bar 1 to the start of bar 2.
Quantize all those notes to eighth notes .
Okay, quantizing everything.
Wait, what is it doing? You need to quantize the MIDI track, you know?
Of course, I'm quantizing exactly that now. I've found the MIDI clip; I'm
now. I've found the MIDI clip; I'm aligning the start of each note to the nearest eighth note.
Done, all notes on the Halion Sonic MIDI track are now on eighth notes, keeping their original duration and pitch.
Doesn't look like it, they still seem a bit off.
Look, the first three notes should be a triad chord together and part of the group, so consequently, they are all triad chords played in staccato eighth notes.
Okay, I'm checking the first few bars and fixing the attack points.
So we played something, and right now it's fixing it for us.
I have a bit of latency when playing, which is why I'm having some trouble, but from the first bar, now every eighth note starts a triad with the three notes aligned, maintaining the staccato.
As I was saying, because I'm using OBS Studio, the computer isn't very powerful , so it gives me some latency, but it's not that the program has latency—in fact , the program works at practically zero latency; it all depends on your sound
card and your computer because this is just a piece of software. Let's hear
what it did for a second.
Let's say at this point I played these triads a bit messily and I wanted to do this.
Can you align these triad chords in a sequence that matches the chords of the backing track?
Okay, I'm checking the backing track and realigning them.
I'm talking about the melody chords, okay?
Understood, I mean the melody, I'm correcting them.
And these triads are chords, put in the right chords, obviously keeping the performance rhythm, basically the performance notes as they are placed.
Okay, I'll keep the rhythm and only change the pitches.
Let's see what it comes up with for this , so we are having an assistant that might give you some suggestions on a note.
It seems to rotate between F, A, E, and G. I would suggest F major, A minor, E
G. I would suggest F major, A minor, E minor, and G major. One chord per bar, changing only the pitch. Should I
proceed with this map? I have already harmonized it this way, keeping the attacks and durations in eighth notes.
That's perfect. I already know it's fine because those were the chords, I remember them now.
Now what, what can we do?
We can work a little magic.
Listen, there are some vocal tracks here okay?
They're sung quite poorly and so on, but to get an idea, could you transform them into vocal tracks sung by a nice, very dance-oriented male voice, maybe even soulful, with lyrics that make sense and
roughly follow this melody that was sung ?
Okay, checking the vocal tracks and options.
So now we're about to do something that would have taken a lot of time because, well, I would have had to write lyrics, sing them this way, and then obviously go find a singer to sing it, record it,
and do the whole recording process.
Starting on the three takes. For the new voice, I'm using DREX, a tenor and R&B dance pop style.
So it chose a woman for me, but not a man; I wanted a male voice, you know?
Yes, I understand. Only male voice. I'm
re-checking and staying on that path; the one I've chosen is an adult male voice.
Two versions with a male voice are in progress, with original English lyrics, a main voice and a second, more subtle one.
Please make sure the melody stays within the chord harmony, okay?
Okay, keeping the melody within the chord harmony.
It might happen.
Inserted now, I'm carefully verifying the chord changes across the whole backing track before tuning the vocals.
It can happen; if they are sung really out of tune, it might interpret a note that is the one, but since they are out of tune, it's not within the chord harmony, so this also needs to be specified.
It executes exactly what you ask, okay?
It doesn't have its own thoughts, wait, it's just a material executor of what you give it, nothing more, nothing less .
I've prepared two parts with DREXELL, male R&B Dance Pop with original English lyrics.
The melody follows the F Major, A Minor , E Minor, and G Major progression. The
third take didn't yield any useful notes ; I've left the original recordings in the project, but muted.
Let's hear what it came up with.
Right now, it's a mess, you can't hear anything, it's all mixed very poorly, so let's make this request.
Can you mix the whole backing track properly so that the vocals are front and center and nice and clear?
All right, I'll prepare a mix with the vocals very much in the foreground.
Also, use her vocal delay on the voice, I like that.
Okay, I’m adding a delay to the vocal.
Valhalla Delay?
Yes, I’m adding a tempo-synced delay to the lead vocal, keeping the repeats slightly lower than the dry vocal so it stays upfront.
Remember that the studio was created, the first version was built specifically for singers and vocals.
Centered with compression and dotted eighth-note delay.
I’m exporting the mix to check the peaks and the headroom on the master.
It even mixes the track for me, or at least gives it a clean-up, then from there you can maybe also go to Mix ready: lead vocal front and center, instruments turned down where they were
masking, second vocal further back, and tempo delay with light repeats.
Peak around -3.6 dBTP without clipping.
Listen to it and let me know if you want even more volume on the vocal or more Okay, I’m noticing the beginning of vocal one has some notes off-beat; can you fix that for me?
And for the secondary vocal, when vocal one isn't there, bring it back up to the front.
Okay, I’m correcting the initial attacks and then I’ll work on the second vocal where the first one is silent.
I’ll check the attacks of the first phrase first, then find the points to bring out the second vocal and set Valhalla Delay as you asked.
So we have a studio assistant doing things for us.
You listen and supervise every single step.
I added a part. The parts are played, they’re mine; I played them, and if I wanted, I could have replayed them. If
another sound came to mind, I’d find it , it would create it for me. In short,
it’s a truly very interesting thing to create with. Obviously, this is the
create with. Obviously, this is the first of many videos.
I’ll also show you how to connect it to your favorite DAW, how many systems there are to start from scratch to create a track, so what’s done by you, what’s suggested by it, some from you, some from it; so all the various steps
that can help you compose or create, or even have the possibility of having an assistant, remote applications connected to the internet to end up controlling Cubase.
Here, instead, with a voice command, everything would be fine. It’s a
fantastic thing, so we have various things.
Now I’ll show you what we can achieve with an idea that’s been sitting there for a year and a half that didn’t inspire me much.
What I asked for is quite complicated, it involves automating and connecting it to the two tracks, so On the two tracks, I’ve already realigned the first phrase to the grid while maintaining the harmony of the
chords.
The second vocal now rises in the pauses of the lead.
Valhalla Delay is on a dotted eighth note. I'll get everything ready for you
note. I'll get everything ready for you to listen to, and then, if you want, we can also set up the volume automation between the two.
Listen, just do the automation for me directly, okay? Just do the automation
directly, okay? Just do the automation so we can get it all done, right?
Yes, I'll do it right now.
And this is simply the studio.
Imagine if I connected it to Ableton, Logic, Cubase, or Studio One.
Guys, it’s something that's science fiction; I mean, in just two years we've gone from making normal music to sci-fi music okay?
So really.
We can listen again or refine the crossover points.
Let's listen again.
Okay, starting playback. I'll restart
from the beginning with both voices and the automation, no loop, so it plays through where the other voice comes in.
Okay, playback started from the beginning, now you'll also hear the second voice in the finale.
We'll bring them down to this level and they seem right to me. Let's see for a moment.
Okay, at this point let's create some magic.
Taking all these parts, okay, that we've created—I authorize you to twist them however you like, or fine, to keep a logic, create an arrangement of about 3 and a half minutes for me.
Use effects, use drum fills, breaks, and add a few pads for when it opens up or other sounds you deem necessary to, especially at the chorus drop, give the track more punch.
All right, I'll plan the arrangement and prepare a timeline of about 3 and a half minutes for you.
Now the magic begins.
I'll let you know in a bit how I extend the sections and the effects.
While it's doing that now, it takes a little while, it takes a minute to do this. Think about how long it takes you
this. Think about how long it takes you to do it.
And it gives you a layout, okay, it won't be final, but you already have a project laid out in a certain way that's already a start to begin working on seriously.
Guys, this is something truly fantastic fantastic fantastic.
I saved a copy of the session and prepared fills, risers, impacts, and a stereo pad that opens the choruses. Now
I'm integrating everything into the 100 -bar timeline and I'll update you on the final structure.
Let's see what it did.
It kept my tracks and even rearranged them.
Look, look at what it's doing.
It's adding things, it's making a complete arrangement from which we can calmly work, okay?
I'm really curious to see what it has.
Arrangement across nine sections on the timeline. In the builds, the beat drops
timeline. In the builds, the beat drops out, then pads, strings, and an impact upon the return of the chorus. Checking
the full render now.
Listen, delete vocal track one, vocal two, and vocal three.
Okay, I'm removing all three right away . I've already found the three original
. I've already found the three original vocals with the silences and am removing them from the project.
Guys, this—this thing is blowing my mind .
I mean, it created a—obviously, maybe you don't take it as gospel, but you already have a structure to start working on, and this, I think until last year, was science fiction.
Kudos to the studio for making this, which is truly something extremely different; anyone else who does something similar is still tied to trial and error to get things done.
Here, instead, it works on your personal stuff, it assists you, gives you a hand , and develops it.
Then, you can jump in, you can play, you can add instruments. Truly an
exceptional thing.
Combine it with your DAW; you can have all the features you usually have in your DAW with this assistant that follows you and integrates perfectly with Cubase, Logic, Ableton, Studio One
—anything that can read a VST.
I'll show you how it interfaces with your DAW.
Okay. The three original vocals have been removed.
The audio files are in the project folder, and I found two clips of the second voice where I need to fix the automation edges to keep the swell in the break and the outro.
I played the various string stabs myself ; it was a loop I would have fixed and cut myself, just testing and hearing how it sounded, working on it, opening and closing it, checking the complexity level.
Look at the de-esser—I mean, on the whole vocal track, it's incredible. Yes
, something like this—I mean, until two years ago, it was unimaginable. This is
a DAW that, used this way—obviously, I’ll show you that doing these sound things, it can make a track from scratch with all tracks separated, sung, and it does everything so even those who don't know how to make music can do it—but
working like this is, in my opinion, an ethical way. Let me know if you think
ethical way. Let me know if you think working like this is an ethical way.
Some parts, maybe a few pads you might have inserted, I see the inserts, but they are all parts played by you and it still respects the chords you decided on .
Done. Vocals 1, 2, and 3 are deleted; the two Drake and Elle vocals remain, and in the choruses, I gave more body to the drums, bass, and pads. I'm saving
the final version. Open the project " When Lights Go Low," 3 minutes and 29 seconds, 100 bars at 115 BPM with fills , stabs, and pads at the start of the choruses.
Let's have a listen, guys.
At this point, obviously, they created a gap that shouldn't have been there, but nothing transcendental, guys, because it would just be a matter of pasting it like this, pasting this part too because they wanted to make the middle empty,
wanted to do something cool, but hey, if we just take a moment.
Problem solved.
They made an exit. Guys, what can I say ? It's a good starting point. There are
? It's a good starting point. There are
these breaks too.
Obviously the drum roll isn't perfectly in time, but I'll show you what it manages to do, for example, very simply .
At this point, I can also recover the original drums because it's in...Load
Groove Agent for me, please, and set a kit named like Stefan, okay?
Alright, I'll try to insert it and look for that kit.
So it's my personal kit, I'll reopen it . At this point, I'll re-record drums,
. At this point, I'll re-record drums, breaks, fills, I'll work on them, I can add, record audio, other synths, other things.
So, starting from something you heard before that was horrifying to listen to , here there's already something more pleasant to hear and to do.
For instance, I’d remove almost all of the second voice underneath and leave it only in some spots, add flourishes, basically, let's start working, well, there you go. Look, it loaded Groove
Agent and found my kit. In this case it’s not playing, but here it is.
My personal kit that is now for the Cubase release.
I loaded Groove Agent SE with the Stefan Prod Set 1 kit.
The pads are ready, plugin open, project saved. Tell me if you want another kit
saved. Tell me if you want another kit or an initial pattern.
So guys, it's something exceptional. It
loaded exactly what I wanted. Usually,
to find this kit I’d have to scroll down , where is it, type, search. It puts it there immediately, so it's a really, really interesting assistant. Alright
guys, tell me what you think of Studio.
For me, it's fabulous.
In the next videos, we'll see how to create a track from scratch, how, as I said before, to have it as an assistant that records you while you're in the studio, gets you started, and then you're left just to create.
We'll also see how to integrate it into your DAW and how to use it continuously with voice synthesis, ChatGPT, and your favorite AI tools.
So, what can I say? If you liked this video, leave a like below, if you like the channel subscribe and remember to hit the bell to know when the next video comes out. So, what can I say? Bye
comes out. So, what can I say? Bye
everyone and see you soon.
Loading video analysis...