Projects 14 min read

THE SATURDAY MACHINE: takes a lot, so choose your why

A 3:56 napkinfilms parody-of-self music video. Italo-disco in D minor, built on top of a chipforge track and loaded up with four autotuned voices, forty German adlibs, a mordant British existentialist chorus ("takes a lot, work work work then you die — so choose your why"), a James Brown tribute shout, a Bach-arpeggio sax solo at 0:52, a disco-lit vending machine with 16-segment beat-synced LEDs, and a Plan 9 bunny who rises from the dispenser slot to wave goodbye while Rocko says "Servus" in the blackout. 132 voice beats. Nine production passes. EBU R128 loudness-normalized. Wears headphones beautifully.

THE SATURDAY MACHINE: takes a lot, so choose your why

THE SATURDAY MACHINE: takes a lot, so choose your why

There is a vending machine humming in an abandoned office park.

Plan 9 the bunny is inside it. Nobody knows yet.

At six o'clock on a Saturday, Coffee Lady walks in and buys a bag of chips. Then Briefcase Guy. Then Hoodie Person. Then Gym Dude. They don't know each other. They are not trying to dance. But the beat gets into them — a toe tap, a head nod, a shoulder sway, then, gradually, all of them are dancing. The machine's LEDs start pulsing on the kick. The chaser runs around the body. The corners flash on bar changes. The ride bell enters. A sax solo lifts at 0:52. A British voice says "takes a lot. takes a lot. gotta keep going. work. work. work. then you die. so choose your why." A German voice — someone's uncle, apparently — says "Ausgezeichnet" at 1:58.

The big drop is at 2:21. The camera shakes on the kick. The final chorus at 3:09 has everybody going. At 3:25 Plan 9 the bunny pops out of the dispenser slot bobbing to the beat. At 3:40 the tape stops. Everyone freezes mid-pose. The lights dim out one by one.

Then — in the blackout — a tiny Plan 9 bunny rises above the machine and waves. A German voice says:

Ausgezeichnet. Auf wiedersehen. Servus. Pfiati nacha. Tschuss.

Then it's over.

Watch THE SATURDAY MACHINE on YouTube. The engine is open source (GPL-3.0-or-later). The film is Creative Commons (CC BY-NC 4.0).

The shape

Fifteen through-composed patterns of Italo-disco in D minor, BPM 122, ~15.74 seconds each, totaling 236 seconds. With the intro and outro bookend, the release is 4:07. Chord progression per pattern: Dm | F | Bb | A7 | Gm | C | F | A7.

  • 0:00 Bookend intro. Chipforge Dm sting (coin click → rising arpeggio → held triad → high bell). At 3.5s a Stranger Things cricket ambience fades in. At 5s the crickets get pitched down an octave, lowpass-filtered, and squashed to silence. Six seconds. The film starts.
  • 0:03 P0 intro_piano. Empty concrete alcove at dusk. Fluorescent ceiling flickers. Vending machine in silhouette. Title card: "THE SATURDAY MACHINE / one coin. one groove." A German voice says Hallo.
  • 0:15 P1 arp_enters. At frame 204 (0:17) the machine LEDs spike to life — the arpeggio has entered, the machine wakes up with it. This is the first visual punch.
  • 0:31 P2 groove_a. Coffee Lady walks in. Buys a bag of chips. Her foot starts tapping while the coin drops.
  • 0:47 P3 melody_intro. Eight fast rhyming lines, one per bar, Samantha at natural speed: "Five past six. Clean and lean. / Quarter drop. Right routine. / Grid lights flash. Bass so deep. / Foot on floor. Can't sleep. / Shoulder sway. Fingers strong. / Beat says move. Can't go wrong. / Saturday hit. Keep it mean. / Machine. Make the scene." The sax solo enters at 0:52 under the rap.
  • 1:03 P4 chorus_a. First full British chorus. Daniel delivers the mordant work-anthem. Behind him a German "Warum?" echoes.
  • 1:19 P5 dj_break. Rapid-fire close-ups: giant shoes tapping, fingers drumming on a thigh, a big-face close-up, the LED strip filling the whole frame. "Vunderbar."
  • 1:34 P6 verse_2. Hoodie Person arrives. German commentary on the new face. "Ze hoodie, ja. Sehr gut."
  • 1:50 P7 melody_var. Sax solo returns in arc + descend phrases. "Die Musik!"
  • 2:06 P8 pre_peak_build. Machine LEDs come alive visibly. Gym Dude arrives, hype-instant dancing. "Ich bin ready. Ach du lieber."
  • 2:21 P9 chorus_b. THE DROP. Full dance. British chorus again. Eve: "Let's go!"
  • 2:37 P10 chorus_b_var. Continuation. Overhead compression. Colors cycling per bar.
  • 2:53 P11 solo_section. Hoodie feature — scaled up 1.35×, others shrunk 0.78×. "Get down with your bad self." (James Brown tribute, Daniel/British voice)
  • 3:09 P12 final_chorus. The biggest moment. Camera shake on every kick. All four characters, full-dance, machine pulsing on every quarter-beat, multiple voice layers calling through the chorus, the Bach arp running 16ths underneath.
  • 3:25 P13 outro_a. Plan 9 bunny pops out of the dispenser slot, bobbing to the beat. Crowd cheers in stick-figure silhouette. "Saturday. Machine."
  • 3:40 P14 outro_tapestop. Tape stop. Everyone freezes. Machine lights dim out. Screen fades. Five German goodbyes layer in: Ausgezeichnet. Auf wiedersehen. Servus. Pfiati nacha. Tschuss. A tiny Plan 9 bunny rises and waves.
  • 3:56 Bookend outro. "A NAPKIN FILMS PRODUCTION" card. Written / composed / animated by Joshua Ayson. Four-bell Dm-triad signoff (D5 / F5 / A5 / D6). CC BY-NC 4.0. License line.

The thesis

The user's brief: "a Weird Al wacky remake of something of my own."

We started from a rendered chipforge Italo-disco track (saturday_machine_score from the April 19 "night of songs" batch — longest, most structurally complete). The film wrapped four autotuned character voices, 40+ German adlibs, a British chorus with an existentialist punchline, a James Brown tribute, a Plan 9 bunny cameo, and a disco-lit vending machine on top.

The core line is the chorus:

Takes a lot. Takes a lot.
Gotta keep going.
Work. Work. Work.
Then you die.
So choose your why.

That last bit is the Viktor-Frankl turn. The song looks like it's going to be about grinding. The punchline is that the grinding is only worth it if you've picked the why. The chorus lands on A7 — the dominant chord, unresolved — so the question stays open into the next cycle. The film doesn't answer it. You get to.

The multi-voice architecture

Six character slugs, each with its own macOS TTS voice, speed, and per-character volume adjustment:

Character Voice Speed Gain Role
fable Samantha 0.88× −1 dB main sung vocals
eve Samantha 1.0× −2 dB fast rhyming, excited shouts
brit Daniel 1.0× +3 dB mordant British chorus
daniel Daniel 1.0× +3 dB emphatic standalone callbacks
german Rocko (de_DE) 1.0× +2 dB comedic Germanic commentary
joshua Tom 0.92× +3 dB measured philosophical overlay

The deeper voices are lifted, the higher voices pulled down, so the mix doesn't turn shrill when everything fires at once. All 132 voice beats route through the same two-stage autotune: a wet layer (heavily pitched to the chord progression) at +1 dB, and a dry layer (raw TTS) at −9 dB for intelligibility.

The wet layer runs through autotune_voice.py --mode continuous --blend 0.85 --max-shift 18 --crossfade-ms 120 with a 120-note per-bar melody. Each word gets pitched to the 3rd of its bar's chord; the 120ms crossfade between bar-notes gives a smooth portamento-like slur from note to note. The result is a sung quality — not just a pitched rap — on any voice that gets more than half a bar of duration.

The Bach layer

Underneath the main score runs saturday_machine_runs_score.py — a secondary chipforge score locked to the main's BPM and chord progression, adding:

  • 16-note arpeggiated pluck_bright runs on every bar — chord-tone ascending-descending arcs in the WTC-Prelude-No.-1 tradition
  • Cowbell on beats 2 & 4 of every bar (the Italo-disco backbeat signature) from P2 onward
  • Ride bell on 8ths during choruses and builds (P4 / P8 / P9 / P10 / P12)
  • Open-hat shimmer on the "and of 4" every bar
  • Corner-flasher crashes at the start of chorus patterns
  • A featured sax solo on lead_expressive at P3 bars 3/5/7 (rising under the fast rap), P7 bars 1/3/5/7 (arc + descend phrases during the melody-var pattern)

Each sax-solo phrase is an 8-note eighth-pattern traversing the chord tones, with a soft-start → peak → soft-end velocity envelope [0.7, 0.8, 0.9, 1.0, 1.0, 0.9, 0.85, 0.75]. That envelope is what makes it sound "played" — the difference between sequenced MIDI and a phrase with breath.

The runs layer sits at −9 dB under the main score with extrastereo=m=2.0 widening. Because the channels already have pan_automation (cowbell right, ride left, arp_clean right, counter-arp left), the widening turns the rhythmic texture into genuine spatial motion. On headphones the ride panning from left to center to right feels like someone walking around you with a stick.

The disco machine

The vending machine is the film's second main character. Its visual behavior is scored to the music as precisely as the drums:

  • 16-segment LED strip across the top — chaser at 16th-note rate (one segment "active" at each sixteenth), FULL flash on every quarter-note kick, palette cycling through 5 color banks per bar.
  • Side vertical LED strips — mirrored chasers (up-left, down-right), active only when the film's ambient glow > 0.35 (roughly P5 onward). Each segment pulses on the kick with its own palette color.
  • 4 corner LED flashers — flash-on-bar-change with full brightness at glow > 0.3.
  • Per-bar display code (A07 / B12 / C07 / …) — updates every bar (was per-pattern in early cuts). "--=" glitch flickers in heavy party mode.
  • Snack grid — the first row of candy-colored tiles tints with the kick when glow > 0.2.
  • 0:17 wake-up — the arp_wake_boost function spikes the machine's overall glow from near-0 to 0.55 over 1.3 seconds exactly when the arp_shimmer channel enters. The machine literally wakes up with the music.
  • Progressive ambient glow — every pattern has a baseline glow level: P0 = 0.00, P2 = 0.22, P4 = 0.55, P7 = 0.72, P9 = 0.95, P12 = 1.00, P14 = 0.30. Crossfades smoothly between patterns. The audience sees color from minute one, escalating all the way through the final chorus.

The post-production

Mix

Four audio stems into the final:

  1. MUSIC — main chipforge score at −3 dB base, pattern-scoped ducking during vocals (12 windows: 0.68× during big choruses, 0.82× during light adlib passages), 4s tail fade, extrastereo=m=1.6.
  2. RUNS — secondary chipforge layer at −9 dB (no ducking — fills vocal space), extrastereo=m=2.0.
  3. VOICE (wet) — autotuned at +1 dB with highpass=80Hz, compand, and 3-tap aecho (80ms / 160ms / 240ms) for a slight-room reverb.
  4. VOICE (dry) — raw TTS at −9 dB with highpass=110 + lowpass=7500 + tighter compand — sits as a barely-audible intelligibility layer under the wet pitched voice.

Everything goes through amix=inputs=4:duration=longest:normalize=0 then loudnorm=I=-14:tp=-1:LRA=7. Final output is −14.3 LUFS integrated, -1 dB true-peak, ready for YouTube/Spotify streaming without further adjustment.

Stereo widening

The extrastereo filters on the music and runs layers give the headphone-listener a genuine spatial field. The chipforge pan_automation already spreads elements across the stereo image; widening amplifies the separation. Listening on headphones you can locate the cowbell (right), the ride (left), the counter-arp (far left), the kick (center), the vocals (center-forward), and the sax solo (center). The song goes around your head during the builds, which is exactly what the user asked for.

Bookend intro

Six seconds total:

  • 0-3.5s: chipforge Dm sting — coin-click metal_sting → rising Dm arpeggio D3-F3-A3-D4 on pluck_bright → Dm triad held on supersaw_pad → D5 bell accent → crash swell.
  • 3.5-5.0s: Stranger Things crickets — classic nocturnal ambience fades in. The cliché that signals "we're outside, it's quiet, something is about to happen."
  • 5.0-5.8s: CRUSHED cricketsasetrate=22050 aresample=44100 lowpass=f=480 afade=out. The crickets get pitched down an octave, hard-lowpassed, and squashed to silence. The joke signals tone: this film is not here to be earnest.
  • 5.8-6.0s: silence. The main song hits.

Bookend outro

Four and a half seconds:

  • The "A NAPKIN FILMS PRODUCTION" card fades in with writer / composer / animator credit and CC BY-NC 4.0 license line below.
  • A four-bell Dm signoff plays — sine waves on D5, F5, A5, D6 (the tonic triad of the song) with long reverb tails. Matches the chipforge intro's D5 bell as bookending motif.

Nine production passes

v1 was music-only finalmix over the raw scene (no vocals). v2 introduced the British chorus and the first attempt at autotune (spit-rap mode — wrong choice, destroyed timing). v3 switched to continuous mode and got the sung quality right. v4 added the German character (Rocko, de_DE). v5 added per-character VOLUMES balancing, the parallel dry voice layer, and lowered the melody by an octave to kill the chipmunk effect. v6 added 15 more vocal beats + ambient glow progressive build + arp-wake moment. v7 added the disco machine upgrade (16-segment LED, side strips, corner flashers, per-bar palettes) + the 16th-note runs layer + cowbell + ride + sax solo. v8 added the chipforge intro sting + branded cards + bookend. v9 added the first-minute density pass (24 voice beats in P0-P3), the 1:35 Hoodie-arrival German reinforcement, the James Brown "get down" line, the farewell sequence with 5 German goodbyes, the Plan 9 bunny wave during the blackout, stereo widening on music + runs, and the cricket-squash intro bit.

Each pass was specific direction: "the German voice is landing funny, give me more of it." "The lights at 2:07 are great, get that feeling earlier in the film." "The chorus needs a British voice saying 'takes a lot, takes a lot, work work work then you die, so choose your why.'" "Arpeggiate some of the voices too." "Make the song feel like it goes around in your head when it's rising." "Add Plan 9 bunny and a little more German at the very end."

The pipeline supports the iteration: manifest-only mode for quick lyric changes, per-stage chipforge re-renders independent of each other, 6-second finalmix loop for mix balance, stem-preserving archives at every pass.

The stack

Same as every Napkin Films production:

  • Animation: Python + PIL line art, 854×480 at 12 fps, 2833 frames core film
  • Main score: ChipForge Italo-disco orchestral — kick_deep, bass_acid, piano_italo, supersaw_lead, arp_shimmer, sub_boom + DJ effects (fx_zap, fx_bubble, fx_tape_stop). D minor, BPM 122.
  • Runs layer score: secondary ChipForge — pluck_bright arpeggios, cowbell, ride_bell, hat_open_shimmer, lead_expressive sax solo, supersaw_pad, crash
  • Intro sting: ChipForge original Dm triad — metal_sting, pluck_bright, supersaw_pad, add_bell, crash
  • Voices: macOS TTS — Samantha (fable, eve), Daniel (brit, daniel), Rocko (german, de_DE), Tom (joshua)
  • Autotune: autotune_voice.py — rubberband-powered, formant-preserving, --mode continuous --blend 0.85 --max-shift 18 --crossfade-ms 120, D-minor chord-tone melody
  • Ambient: Stranger Things cricket chirp (ChipForge) + inline-generated Dm-triad sine-bell signoff
  • Mix: FFmpeg with 4-stem amix, pattern-scoped music ducking, loudnorm for EBU R128 perceived-loudness normalization, extrastereo widening for headphone depth
  • Bookend: PIL-rendered title cards with disco navy-magenta palette + LED-strip motif, ffmpeg concat with fade-aligned video transitions
  • Direction: Claude Code, Opus 4.7, agent mode

No GPU. No stock footage. No licensed samples.

Credits

Written, directed, composed, animated, voiced, and produced by Joshua Ayson in collaboration with AI. Made by Organic Arts LLC, Nevada.

Plan 9 the bunny voiced by silence. Customer voices via autotuned macOS TTS. The German voice is Rocko; the British voice is Daniel; the two Samanthas are Eve and Fable; Tom plays the quiet Joshua overlay.

The line

Takes a lot. Takes a lot.

Gotta keep going.

Work. Work. Work.

Then you die.

So choose your why.

— The Saturday Machine


Related work

THE SATURDAY MACHINE sits in the Napkin Films dance-music-video line, distinct from the quieter meditations like GRIEF WORK or TRANSMISSION. This one earns its place by being genuinely funny — the German voice is an ongoing joke, the British chorus is a stoic joke, the Plan 9 bunny farewell is a tender joke.

  • GRIEF WORK — 2:47 tidal ambient, the bunny releases a warm orb and rises to air
  • TRANSMISSION — 3:01 meditation on self-as-universe, "hello, me"
  • ENCODED — consciousness-as-signal
  • UPTIME 32 — 32 bits, still processing

If you want the origin of the Napkin Films stack, Four Films From Code covers how the architecture works and why constraint is the feature. The Agentic Development post explains the agent-mode workflow that made this film possible in nine passes.

Production insights: Full lessons-learned in docs/insights/saturday-machine-insights.md — what worked, what failed, what's reusable.

License. THE SATURDAY MACHINE is licensed under Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0). Share and adapt with attribution to "Organic Arts LLC" and a link to the original, non-commercial use only. Engine code is GPL-3.0-or-later. ElevenLabs / macOS voice audio is licensed content and is not redistributed. Contact: j@organicartsllc.com


Produced with Napkin Films and ChipForge, open source tools built by Joshua Ayson and AI agents at Organic Arts LLC.