Incoherent innovative is the best it can do.

And even this is an illusion. None of the current models should be called “AI,” they are merely highly capable autocomplete and operate only by repeatedly answering the question: “what token should come next?” In the text models a token is a few characters and in the music ones a few milliseconds of audio. Whatever illusion of creativity they fool you into thinking they have is merely randomness introduced so that they don’t always produce exactly the same output (that is, so they sometimes pick an output that is not exactly the most likely next one based on the training data).

I make trance and since I’m not able to imagine complex sounds in my head I find it helpful to mix as I go, including mixing into a simple mastering chain of compression, saturation, and a limiter.

I have been trying to release a track every two weeks since late 2023 and I do everything myself. I’m sure it shows, but that’s where I am in my music journey at the moment.

I have been trying to release a track every two weeks since late 2023 and I do everything myself. I’m sure it shows, but that’s where I am in my music journey at the moment.

No. None of the current models should be called “AI,” they are merely highly capable autocomplete and operate only by repeatedly answering the question: “what token should come next?” In the text models a token is a few characters and in the music ones a few milliseconds of audio. Whatever illusion of creativity they fool you into thinking they have is merely randomness introduced so that they don’t always produce exactly the same output (that is, so they sometimes pick an output that is not exactly the most likely next one based on the training data).

If they survive the copyright war, these models will find applications in places like marketing. They will not replace actual art production where humans are connecting with other humans.

Making trance, for me it’s:

1. Kick
2. Bass
3. Chords or melody
4. Other one from 3

Why not just take one of those as a reference and do what they do but with your own sounds? That way you know you’re adding enough, since they’ve already done it.

My man punching way above his weight class.

Inspiration: do the work every day.

Motivation: see Inspiration.

1. The chords around 1:40 have the Some Chords sound, nice work. They could maybe be detuned just a bit less if you really want to lean into that sound.
2. The other thing I think of when I think of Deadmau5 is really nice (but softish?) plucks, and there really weren’t any in this track that gave me that vibe.
3. Composition and arrangement wise, this seems like at least two tracks mixed together around 3:30, with maybe a third one coming in near the end. Pick one of these and fill the whole track out with it, and you can make a second track with the other part.

A few comments:

1. There is a lot of crackling in the first part: if that’s intentional, OK, but it’s quite noticeable so it might be a bit loud.
2. The pitch swept “bwom” sound is a little out of place to me: maybe it could be detuned more or otherwise moved back in the mix.
3. The buildup feels the same level as the drop to me, but I’m missing a lot of sub bass listening on my phone so that might be it.

If it helps, try to remember that you’ll eventually release something worse than whatever it is you have ready to go right now 😂

Source: I released a track last year that is much worse than anything else I’ve made. It was not my first released track 😬

I just tested this with a saw wave and found that up to about 60% drive it's pretty linear, though not exactly as it doesn't 100% null. Beyond that the waveform does start to very obviously round off and saturation is occurring. In further experiments another thing that I'm hearing is that even in the mostly-linear range this gives an alternative control for the distortion effect since it acts as an "input trim" that seems to make the distortion behave differently than increasing the distortion's drive.

Need somethin’ blown up?!

Might be time to pull the trigger on this one. I see it recommended all the time.

I do use Live so will dig into this for some inspiration.

I have it but went back to Serum after a short while, will take another look!

Windows 11 appears to prefer to schedule “background” tasks to the efficiency cores present in 12th gen and later Intel processors. In the Task Manager you can set the CPU affinity to just the performance cores and their logical cores to work around this. Right click on Ableton.exe and select CPU Affinity. Then select just the first 2N CPUs, where N is the number of performance cores your CPU has. For example, if you have eight cores, deselect everything starting with CPU16.

As others have said, ripping from YouTube is not going to give you the correct levels. This track is available on beatport so you could buy the lossless WAV or AIFF there to use as a reference.

One I’ve been referencing recently is Spectral by Allen Watts. It hits -4.6 short term in the second drop at 2:20 or so.

It’s Friday night. I have no date. It’s just me, a two-liter bottle of Shasta, and my all-Rush mix tape. Let’s do this.

Use something like mailchimp (free to start) instead of a regular email service.

1 and it’s not close.

What genre? The modern techy trance I’m referencing is generally 0.0 or -0.3 dB true peak, getting WAV/AIFF from beatport. These tracks generally hit about -4.5 dB LUFS short term and maybe just above -4 momentary in the drops.