There are all kinds of dependencies in life. Us software developers have a few extras that “regular” people don’t, specifically software dependencies.
I was reviewing talk proposals for ConFoo and ran into this video by Darcy Clarke, which covers the terrifying differences in dependency resolution between Javascript package managers. This gave me the initial idea for the song, in particular the line “All of my friends bring all of their friends and all of yours bring theirs too“. Inspiration comes from the oddest places, and Darcy might be a bit surprised by this!
Of course we also have the same dependencies as everyone else – friends, family, lovers, colleagues, medications, recreational habits (chemical or otherwise), income, a civilised society, a functioning economy, breathable air, and so on. This song is mostly about the software thing, but the other stuff can’t help leaking in, as reality is wont to do.
In software, a sometimes tricky thing to deal with is a circular dependency, where something depends on something else, that depends back on the thing you started with, and it all gets stuck going in circles. However, in a non-software context, you might actually like the idea of a mutual, trusting dependency.
I did write a verse that made a play on dependency injection, but I couldn’t find anywhere to stick it in.
This is probably the quickest I’ve written a song, taking just a couple of days to get the bulk of it in shape, and to remind myself of how to play my guitar, but it’s a relatively big mix (for me), with more vocals than usual, which took a while to nail down.
I decided to forego my favoured synthetic vocalist for the main vocals (though I’m still using it for backing), and have a crack at singing it myself, which not I’ve dared do since Tailwind, but it came out OK, partly thanks to the wonders of Logic’s Flex Pitch editor. Right from the start, and obvious from the lyrics, this song was intended to be a duet, so I enlisted the help of a friend’s daughter who happens to like singing – thank you Greta! Greta had never done any proper recording before, and learned some mic technique, the joys of loop recording, comping, and using the pitch editor to try out harmonies.
As in my previous song, Uncomfortable, I gave Logic’s virtual players a workout, playing drums, bass, and a couple of keyboard parts. I played the guitar parts, apart from the epic lead solo, which was played by Wassim Rahmani from Fiverr, who also played on Uncomfortable. I also used Claude again to help out with chord progressions; I would never have guessed at using Em like that, nor the final, dangling Am7(9).
Claude’s idea of an E7 chord
Amongst all this, Claude created the first musical hallucination I’ve seen from an LLM; I asked it to draw a chord box for the E7 chord, and it generated this:
While beautifully symmetrical, this is not an E7 chord, or (apparently) any other chord (try playing it, in either of the two interpretations of this diagram; they both sound horrible!), and nor does my guitar have 7 strings. It also stated that the “o” above the box indicates that these strings should not be played; that’s not correct – the “o” denotes an open, unfretted string; an unplayed string is usually denoted using an “x”. It also provided a text description of how to play the chord, but this had no correlation with the diagram, and was also completely wrong. Further evidence that AI is not about to conquer the universe.
The instruments are Logic’s “Smash” drum kit, Logic’s “Rock” modelled bass, a sampled marimba, an analogue synth patch in “RetroSyn”, and “Disco Strings” from the Studio Strings instrument. Logic’s built-in instruments are really good.
[Verse] There’s too many things that I don’t control; I really don’t know which way to turn. Got an overwhelming urge to get rid of it all; have the feeling I might crash and burn
[Verse] I asked my friend for advice, she said “Well, it depends. Do you really want to go that way? I don’t think you’ll like how it ends”
[Chorus] So many things that depend on me so many that depend on you If we get it all together we can cut them down, until it’s just me and you
They say to never trust a stranger, yet here we all are, handing over keys to the kingdom. Do we really need to go that far?
[Verse] All of my friends bring all of their friends and all of yours bring theirs too. (there’s just too many people here) They’re in such a mess that nobody’s sure exactly what they’re meant to do
[Break]
[Chorus] So many things that depend on me so many that depend on you If we get it all together we can cut them down, until it’s just me and you
They say to never trust a stranger, yet here we all are, handing over keys to the kingdom. Do we really want to go that far?
[Outro] So many things that depend on me so many that depend on you When all I want is a circular dependency with you
As always, I really appreciate reposts of links to my songs on here or on BandCamp.
I find it strange to be able to say that I’ve now created several songs that use a synthetic vocalist. This is a somewhat weird concept, but it’s right at the bleeding edge of music technology. We’ve had voice synthesis for years – I remember using a Texas Instruments “Speak & Spell” when I was small in the 1970s, and it’s gradually got better ever since. The first time I ever heard a computer trying to sing (I’m not counting HAL singing “Daisy, Daisy” in “2001”) was in a Mac OS app called VocalWriter, released in 1998, which automated the parameter tweaking abilities of Apple’s stock voice synthesis engine to be able to alter pitch and time well enough for it to be able to sing arbitrary songs from text input. It still sounded like a computer though. A much better “robot singer”, released in 2004, was Vocaloid, but even then, it still sounded like a computer. A Japanese software singer called UTAU, created in 2008, was released under an open source license, and this (apparently) formed the basis of Dreamtonics’ Synthesizer V (SV), which is what I’ve been using. SV finally crosses the threshold of having people believe it’s a real singer.
The entry of my song in the 2024 Fedivision song contest sparked quite a bit of interest. I posted a thread about it on Mastodon, and I wanted to preserve that here too. One commenter said “I thought it was a real person 😅” – which is of course the whole point of the exercise!
SV works standalone, or as a plugin for digital audio workstations (DAWs) such as Apple’s Logic Pro, or Steinberg’s Cubase, and is used much like using any other software instrument. It doesn’t sing automatically; you have to input pitch, timing, and words. Words are split into phonemes via a dictionary, and you can split or extend them across notes, all manually.
Synthesizer V’s piano roll editor
In this “piano roll” editor you can see the original words inside each green note block, the phonemes they have mapped to appear above each note, an audio waveform display below, and the white pitch curve (which can be redrawn manually) that SV has generated from the note and word inputs. You would never guess that’s what singing pitch looks like!
For each note, you have control over emphasis and duration of each phoneme within a word, as well as vibrato on the whole note. This shot shows the controls for the three phonemes in the first word, “we’re”, which are “w”, “iy”, “r”:
The SV parameters available for an individual note, here made up of three separate phonemes
This note information is then passed onto the voice itself. The voice is loaded into SV as an external database resource (Dreamtonics sells numerous voice databases); I have the one called “Solaria”. Solaria is modelled on a real person: singer Emma Rowley; it’s not an invented female voice that some faceless LLM might create from stolen resources. You have a great deal of control over the voice, with lots of style options (here showing the “soft” and “airy” modes activated). Different voice databases can have different axes of variation like these; for example a male voice might have a “growly” slider:
Synthesizer V’s voice parameters panel
There are lots of other parameters, but most interestingly tension (how stressed it sounds, from harsh and scratchy, to soft and smooth), and breathiness (literally air and breath noise). The gender slider (how woke is that??) is more of a harmonic bias between chipmunk and Groot tones, but the Solaria voice sounds a bit childish at 0, so I’ve biased it in the “male” direction.
The voice parameters can’t be varied over time, but you can have multiple subtracks within the SV editor, each with different settings, including level and pan, all of which turn up pre-mixed as a single (stereo) channel in your DAW’s track:
Multiple tracks in the SV editor
In my Fedivision song, I used one subtrack for verses, and another for chorus, the chorus one using less breathiness and trading “soft” mode for some “passionate” to make it sound sharper and clearer.
This is still all quite manually controlled though – just like a piano doesn’t play things by itself, you need to drive this vocalist in the right way to make it sound right.
Since the AI boom, numerous other ways of getting synthetic singing have appeared, for example complete song generation by Udio is very impressive, but it’s hard to make it do exactly what you intended; a bit like using ChatGPT. Audimee has a much more useful offering – re-singing existing vocal lines in a different voice. This is great for making harmonies, shifting styles, but only really works well if you already have a good vocal line to start with – and that happens to be something that SV is very good at creating. I’ve only played a little with Audimee; it’s very impressive, but lacks the expressive abilities of SV; voices have little variation in style, emotion, and emphasis, and as a result seem a little flat when used for more than a couple of bars at a time. Dreamtonics have a new product called VocoFlex that promises to do the same kind of thing as Audimee, but in real time.
All this is just progress; we will no doubt see incremental improvements and occasional revolutions, and I look forward to being able to play with it all!
This song takes the perspective of a woman receiving unwanted attention at work, how this makes her feel uncomfortable, and how her friends support her in dealing with it. My intention is to draw attention to the fact that this happens too much, that we (men) should call out this behaviour, and help when we can. In addition to my small attempt at empathy, this is also something of a technical experiment in songwriting which is quite a departure for me. Hit play and keep reading!
Apple’s Logic Pro 11 added “session players”, virtual musicians that can decide what notes to play, and I thought I’d have a go at using them. There is a long history of generative virtual musicians, like the algorithmic “M” by Intelligent Music (released in 1987, before search engines existed…), and the pattern-based “Band in a box” (released in 1990, and still going!). Logic’s session players are very similar in style to the “drummer” tracks that were introduced in Logic Pro X in 2013 (I feel old!), and that I’ve used extensively, but add a melodic and harmonic layer. This means that they have to be told what chords to use, and so Logic Pro also added a “Chord track” to do exactly that. You can also have separate chord sequences for individual regions within a track so different players can play different chords at the same time, instead of all sharing the global set, but I’ve not used that here.
Now we come to a bigger problem: which chords to play? Logic has an overall key signature, and given that, it can suggest popular chord progressions for the chord track, like i-III-VI-VII, which it uses to pick chords from the current key. I tried messing around with different keys and progressions, and came up with something I liked (ending up in D minor). Then I thought I would try playing with time signatures, as (unlike me) these players don’t get confused by things that are not in 4/4, and ended up with a nice rolling pattern in 7/8, which is also slightly “uncomfortable” and off-balance, which I thought would go well with the theme of the song.
The built-in chord progressions are very limited, and unfortunately, so is my music theory. So I turned to Anthropic’s Claude LLM for some suggestions on how to continue my song:
Claude prompt and response providing chorus and bridge chord progressions
I experimented with variations on this prompt, plugging the chords into Logic until I was happy with the result; you can see the same chord names appearing in the chord track:
Logic’s chord track
It wouldn’t surprise me if Apple goes much further in this direction. With the local AI engine scheduled for macOS 15 and iOS 18, it would seem an obvious target. Music theory rules are far simpler and mathematical than written language, and thus a great candidate for a smaller (when compared to gigantic online models like Claude) learning engine to produce good results in a constrained, lower-resource local environment. The session players already achieve great results without any such AI engine.
I’m not going over the top with these players; I’m using one each of the bass and keyboard players, with very conventional electric bass and classic rock organ instruments, along with liberal use of their complexity, swing, humanisation, and simulation features. I think the bass in particular sounds great, really believable when it’s by itself, like at the very beginning.
A bass player instrument region and some of its settings
With all that done, I had a decent backing track, and it was time for lyrics and vocals.
Logic’s arrange window for this song, also showing parameter automation
For years I’ve seen endless stories of women (especially in tech) being subjected to miserable, denigrating, sexist, misogynistic bullshit and unwanted attention. This is something I’ve never had to deal with myself (being a stereotypical cis, middle-aged white man), but I know it happens, and I wanted to do/say something positive that I might not have done/said otherwise. I’m not looking to usurp womens’ ability to stand up for themselves, just to show some empathy. To that end, my inexperience may have resulted in something overly naïve and optimistic, however, I didn’t want to write a negative or pessimistic song, and I doubt many want to listen to one either!
To try to ensure that I wasn’t treading on anyone by doing this, I asked several women to read this and listen to the song before I promoted it anywhere. I received generally positive feedback, from “No red flags” and “…release it. It’s good”, “that’s really cool”, “love the voice”, “We like this”, “it’s a great song to have written. Raising awareness of these kinds of problems helps, [no matter who] raises them”, to understandable concerns about big AI’s inherently exploitative nature, and my presumption of speaking for women. Songs written from a different gender’s perspective are not at all unusual: almost everything by Abba was written by men from a woman’s perspective (and not all of it flattering!); Aerosmith’s “Don’t want to miss a thing” was written by a woman from a man’s. If I was using this song to be negative about women, or to displace someone else’s efforts, I would expect severe and justified pushback, but I’m not doing that, and I think the song has a good message regardless of who wrote it. In the other direction, I’ve not had any complaints from men about how my representation of their actions might be considered unfair – but I wasn’t expecting that anyway. Ultimately, it’s just a song, with a little empathy, and a big dose of technology.
The lyrics are a little cheesy in places (yes, Claude is very helpful for finding rhymes!), and I use “attention” too much, but I hope they do the job of conveying the discomfort of unwanted attention, and the relief when it goes away and you have a support network to fall back on. Some might dismiss this as “woke”, but they are likely the ones that this doesn’t happen to, and I didn’t write this song for them.
[Verse] I’m minding my own business; Maybe you should try that too. Consider that maybe, just this one time, it’s really not about you.
[Chorus] I’m not seeking your attention, just trying to get through my day. You’re gonna make me uncomfortable if you carry on this way.
[Verse] With all this unwanted attention I could be flattered some other time. Your persistence won't change my mind; the decision’s only mine
[Chorus] I’m not seeking your attention, just trying to get through my day. You’re gonna make me uncomfortable if you carry on this way.
Your inappropriate advances, just confirm what I should’ve known: You’re making me uncomfortable; won’t you leave me alone?
[Break]
[Verse] No, I don’t need your permission to choose the people that I like. My friends respect my life choices, and support me day and night.
[Chorus] Now I’ve avoided your attention my friends say they can see I’m feeling so much more comfortable in their good company
For the vocals, I turned again to the still-astonishing Synthesizer V Studio Pro (SV) from Dreamtonics, with its excellent Solaria voice. I’ve written about SV before, if you’d like to know more about how SV works. One specific point that I’ve been asked about is that SV is not “AI” in the ChatGPT sense; its voice models are created from specific individuals who are paid and credited for their work, not by slurping monstrous quantities of stolen, uncredited, unpaid input and generating unethical approximations as most public LLMs do (and I recognise this includes Claude). It’s not a huge sample library like many virtual instruments, but a tuned algorithmic approach, more like what Pianoteq does for piano synthesis, though not going as far back to first principles. I don’t see any ethical issues with using synthetic voices created this way; artistic concerns are are a different matter, though so long as they continue to sing better than me, I will keep using them!
SV still has problems integrating with Logic – its tempo sync is hopeless, so I’ve had to stick to a fixed 100bpm. It lacks support for other time signatures, so bar numbers and loop points are never in sync when working in 7/8, and note placement involves a fair bit of trial and error. Despite all that, it’s still quite workable, and sounds great. I made a particular point of adding breath sounds, which really adds to the realism. I did the backing vocals with SV too, and for some reason I found them quite tricky this time, much more difficult than in my other songs, probably because this song is more complex than my others.
At this point I had a workable song, but it needed something more than a semi-automated backing track. I attempted to play some guitar parts myself, but I found the 7/8 timing difficult (I did find it’s easier counting 1-2-3-4, 1-2-3 than 1-2-3-4-5-6-7), and I’m no good at playing B♭ bar chords cleanly, so instead I found the amazing Wassim Rahmani on Fiverr.com and commissioned him to create rhythm and lead guitar tracks. He did a great job; there’s some lovely, liquid playing here, Clapton-ish in places, and he created a much more satisfying ending than I started with.
After that I needed to mix and master, as usual trying to keep all these parts from treading on each other, dialling back the reverb to avoid muddiness, checking stereo spread, not going overboard on compression or levels. I used Logic’s new Mastering Assistant, which gives nice high-level control over global EQ and dynamics. I’ve previously used Logic’s mastering presets (a less sophisticated approach), and IK Media’s Ozone, but the assistant is easy to use and sounds great.
Logic’s Mastering Assistant at work
In all I’m pretty happy with this song. It’s got a good message, and is musically more ambitious than other things I’ve tried, thanks to some artificial assistance. It still feels like an engaging, creative process, but the balance has shifted slightly higher, towards directing and producing, and away from choosing or playing individual notes. I also see clear parallels between what I asked of Claude, and what Wassim did for me, and I’m not sure how I feel about that.
As always, I really appreciate reposts of links to my songs on here or on BandCamp.
This book is the latest incarnation of things I’ve produced relating to HTTP/3. The first was a conference talk I first gave at ConFoo 2024 in Montreal. That went down pretty well, and I was asked to do the same talk at the International PHP Conference 2024 in Berlin. The room was so packed for that talk that they have asked me back to do the same task again at IPC Munich 2024 later this year!
S&S Media, the German company behind IPC, publishes lots of magazines for developers, and I was approached by their representative in Berlin about writing an article on this subject. So after a little negotiation (particularly about rights!), I sat down to turn my talk into an article. This was made quite easy given the framework I had from my talk, but I was pretty pleased and surprised to be able to crank out a decent 4,000 word article in a bit over 2 hours. It’s not been published yet, but I’ll amend this post when it has.
I’d enjoyed writing this article so I wondered about expanding it into a short book. Asking around on Mastodon suggested that the easiest way to go was self-publishing on Leanpub; I didn’t want to get pulled into writing some monster tome for a publisher.
It took about 2 days of writing, plus another day or so learning the ins and outs of the Markua markup format, tweaking, proofreading (thanks to my mum!), drawing diagrams, and building a site to demonstrate the difference in network activity across different HTTP versions.
A week on from my first thoughts of doing this, I’m very happy to find myself with a freshly published 54-page, 12,000-word book! I’ve of course publicised it across all my social media, but need to put a bit more effort into marketing, perhaps look into Leanpub’s translation services. Anyway, we will see how it goes, but it’s been a good experience so far.
