Synfire 2.0 Preview

There is a new build 1.10.11 for the Mac listed under 'Patches' in user accounts, but that's not really an official release. We merely fixed slow graphics output on very large screens.

I must admit this is taking longer than expected, but here's a first look already into the revamped Audio & MIDI Setup. Many of the changes here are indicative of a lot of things to expect next.

To throw in a few other details, Audio Engine 4.0 adds support for VST3 and, among other things, there will be an additional MIDI Effects Drone that loads into your DAW to feed it with MIDI streamed from Synfire directly (to play Logic X internal instruments, for example). 

In the next video I will show some of the new Library, which will take a very prominent role in version 2 (hint: it's also generating phrases). In the third video to come, we'll get a first look at the new arrange window.

Right in this moment, I'm in the final steps of refining exciting stuff with KIM Factories that I can't wait to show you. This is work still red hot in the making, but I hope to get it presentable in 1-2 days. This generative stuff is so groovy and fun, I often find myself jumping out of the seat without noticing. That's probably a good sign ;-)

Still working on the video. Crashes still pop up often and lead into a rabbit hole of code changes that are difficult to predict.

Here's a few screenshots for today, at least, so you can have a quick look already and imagine what's to come.

The images are listed below (full resolution).

Will this work with classical and film style orchestrations?

Yes. The idea is to generate rhythmically consistent fragments that you can drag & drop from your generated libraries into a counterpoint arrangement to compose build-ups and other parts. Most pop/rock/dance users have little grasp of ostinato, crescendo, accelerando, staccato, legato, etc. but might embrace these terms once they see a strong use case ;-)

I have an experimental feature that extracts rules from existing phrases. Nothing perfect, though. What you can do is re-use the rhythm of any phrase and generate different melodies based on that.

Eventually it might make sense to expose the new KIM language to end users, so everyone can write their own factories, but at this time there is no time to even think of writing another 600 page manual ...

KIM is a full-blown rules based programming language, not unlike Prolog. No chance anyone learns this without proper instruction and documentation. I'll need some raw documentation myself, or in a year or so I'll be unable to write factories. But there are more obstacles to exposing a development environment in Synfire with syntax coloring, parser, compiler, debugger, etc. I'm glad it works for now.

The 2.0 user manual will be online, based on a DITA database, so it can be maintained and extended as needed.

Sorry, no WAV tracks yet. Mostly because the Audio Engine has no GUI for that yet.

I'm sorry you still have to wait. Just wanted to let you know things are going extremely well here. During the past several weeks, I went down a number of rabbit holes, many more than anticipated, along endless fractal lines, entering into a wonderland that is beyond my original imagination. This will be HUGE. I don't have a penchant for hyperbole, but this really works out so well, so musically sound, I'm sure it will knock your socks off!

Cognitone took a big risk investing in uncharted territory that is unpredictably costly to explore. The KIM language had only vague contours at first, and it could not be evaluated without actually implementing it to 90%. We put all eggs in one basket, for more than three years now, so it is a big relief that my intuition finally proved right.

As always, the last 5% take the most time, but we're almost there now. There is a lot of room for future enhancement, but I think it is important to make a cut at some point and let real world users get their hands dirty. Can't wait to see all of you extract a hell of fresh music out of it ;-)

We're getting closer by the day ...

Ok, I could no longer wait. Had to make this short 5 minutes preview for you right now. It's really only a quick glance for first impressions. No voice over yet, because it's late here already and I really have to get a break ;-)

At the beginning, I demonstrate how rhythm is generated, which is the basis of phrase synthesis. After a brief look into generating harmony, I do a few quick clicks through different phrases that are 100% KIM output. I'll go into more detail in a follow-up video later.

A KIM Factory basically works like a modular synth, only that it generates phrases instead of sound. After a while you get used to the controls and operate them much like ADSR, MOD, WAVEFORM etc. Once you have a setting that generates a good phrase, you can expect every 3rd to 5th click to produce a wonderful variant of it full of surprises and inspiration. After a hundred clicks or so, your head starts spinning and it begins to dawn on you that you are about to get lost in music. That's when the "Keep Only Favorites" function comes in handy. It deletes everything but the phrases you liked. A big relief that not only frees up megabytes of memory, but also hands you a collection of unique phrases that have never seen this planet before ...

Happy Holidays!

Synfire does parallel computing already, albeit not on separate CPU cores. KIM doesn't need it. Smalltalk is an extremely fine-grained and huge network of objects. It isn't well suited for parallel number crunching. Since literally everything is an object, including classes, compilers and code (Oh, and nil is an instance of class UndefinedObject, so in Smalltalk even the null pointer responds to messages), you would have to deal with very complex synchronization challenges. Aside from bulk import of MIDI files, there is little use for multi-core. That the Audio Engine runs on a separate core is already very helpful.

Smalltalk's strengths are with knowledge representation, inferencing, huge network-like data structures, generic algorithms, non-local returns (backtracking), and building compilers/interpreters for new languages like KIM. Stuff other languages do not support. Much of this extreme dynamism would even be slower if implemented in C++ due to its memory management.

(sorry for the technical gobbledygook, but as some of you are IT pros, everyone else please forgive me)

Do you plan to be able to determine settings for the variation of the velocity when creating rhythms?

Yes, that's already done. Velocity has huge influence on a phrase.

with the VST 3 support, will a single device be able to address multiple ports of the VST 3 instruments

Good point, I'll have to check that.

I run multiple instances of the same app

Absolutely. That could be one way of doing long running heavy-duty work like mass MIDI file import. It's still a challenge to do right.

Ok, so now this will be the last preview video this year:

There's one mistake in it that bugs me: Of course, if you want a basso ostinato, you'll choose blue (horizontal) symbols and push them to lower playing range. In the video, I left it with bass symbols, which however follow the chord progression.

As we are at it, for lack of a better term, Ostinato is used very broadly here, meaning the repeated occurrence of the same pitch. Although, depending on symbol type, this will not always lead to the same pitch.

Hopefully everyone of us will have a better 2021. I'm certainly looking forward!