Agent Mode: Plan 9 and OG Bobby Johnson Rap Battle in Python (175s, All Code)
A cyborg bunny and a russet potato rap about agentic coding over a 15-channel festival-trap beat. 175 seconds. Five WhatsApp voice memos chopped into DJ loops. Everything you hear is Python.
Agent Mode: Plan 9 Meets OG Bobby Johnson (175s, All Python)
Plan 9 the cyborg bunny shares a microphone with OG Bobby Johnson, a russet potato in a backwards cap and a gold chain. They trade verses about agentic coding and street loyalty over an 88-BPM D-minor festival-trap beat. Five WhatsApp voice memos from real friends become DJ loops. Two custom ElevenLabs voices carry the rap. A scrolling poetic manifesto crawls across the screen for 175 seconds.
Watch on YouTube: https://youtu.be/bMBzJusG78w
The brief started with five WhatsApp voice memos: me freestyling about agentic coding ("I don't type, I direct"), my friend doing a verse about loyalty and his boy Jax. The ask was to turn these into a real song with Plan 9 the cyborg bunny and a new character called OG Bobby Johnson, who needed to be funny and come from the kitchen.
So OG Bobby Johnson became a russet potato. Backwards cap. Gold chain with a $ pendant. Eyes built into his body (the spuds). Tiny stick arms holding a microphone. Above all, the song needed to actually sound good: silly characters, real production.
Here is how it came together across ten passes.
The Song
ChipForge instrumental at 88 BPM in D minor: classic trap tempo, a 4-bar Dm-Gm-Dm-Am vamp that loops through the song. Fifteen channels in the final mix: layered 808 and sub bass for floor weight, kick with a punchy attack layer, tight rap snare on 3, 16th-note hat with 32nd-roll fills into the DROP, filtered piano stab loop, wobble bass for the BREAK and DROP, an atmospheric Eno-voice pad underneath everything, an Illenium-style classical hook melody on the bright lead, an Egyptian-Phrygian counter-hook on the snake-charmer scale (where Snoop swing rhythm meets the Nile), a fast 16th-note arpeggio through DROP with an octave-up shimmer voice doubling it, and FX risers building into every section transition.
The master bus runs three engine improvements that landed during the prior plan9_emerge production: oversampled true-peak limiter, parallel air exciter for genuine 6-18 kHz harmonic content, and a mud-zone bell cut at 320 Hz. Mix scorecard: 12 PASS, 1 WARN, 0 FAIL.
The Voice Cast
Three custom ElevenLabs voices designed specifically for this universe:
OG Glenda Bunny (Plan 9) carries fast spit-rap and any future film where Plan 9 speaks. OG Bobby Johnson (the russet potato) delivers husky tongue-in-cheek lines. The Creator (my own cloned voice) appears once at the outro for a fourth-wall cameo: "...OK that was something." Bella provides female back-up ad-libs: "Yeah!" "Let's go!" "Agent mode!"
These voice IDs are documented as canonical mappings. Any future Napkin film featuring these characters uses the same IDs.
The Sample Chops
Five real WhatsApp voice memos get chopped, time-stretched 3-5% via rubberband, autotuned to D minor pentatonic, and dropped DJ-style at HOOK entries. Bobby's "I support..." 4-bar phrase loops six times through HOOK 1 starting on bar 10. My "I don't type I direct" loops five times through HOOK 2 starting on bar 22. Synchronized "Agent Moboon" and "I support Jax" chops fire together at the bar-42 DROP for the big DJ-style synchronized drop moment.
The autotune dial got pulled back over multiple iterations: the samples need to stay intelligible. Final settings: speed 1.04-1.05, transpose -1 semitone, spit 0.65, blend 0.55, max-shift 6 semitones to D minor pentatonic. Light T-Pain feel, not the heavy chipmunk treatment.
The ElevenLabs Filler Raps
50+ ElevenLabs filler lines fill the gaps so no second of the 175 seconds is silent when characters' mouths are visible. Bar-aligned to BPM 88. Six languages of Russian goodbyes from Plan 9 at different moments through the film: Plan 9 comes from a different galaxy and misses home.
The Visual Feast
PIL stick-figure animation at 12 fps. Plan 9 stage left with cyan chip glowing on his forehead, OG Bobby Johnson stage right with the backwards cap and gold chain. Three kitchen backup dancers in the corners: egg, lettuce, and carrot, bouncing on the kick. EQ visualizer bars on the platform front edge during HOOKS and DROP. Falling neon confetti during HOOKS and DROP. Audience silhouettes across the bottom raising arms on every kick. RGB lasers sweeping during the DROP. Per-section camera push-pull (push toward Plan 9 in verses 1 and 3, push toward Bobby in verse 2, pull back for BREAK, kick-shake during DROP). Digital glitch slices on the heaviest DROP kicks.
OG Bobby winks at Plan 9 every 8 bars. His eyes track Plan 9 on the snare beats. The egg dancer color-flashes pink on every kick. The lettuce dancer's leaves flutter on the kick. Personality details keep the corners alive.
The Scrolling Manifesto
The previous lyric callout style (single words like "AGENT MODE" or "I DIRECT") got replaced with a continuously scrolling poetic manifesto written specifically for this film. It runs in two parallel streams at different speeds across the upper-third for the entire 175 seconds. Brain-spin effect: you can't quite catch every word, but every word does eventually scroll past, and individual phrases stay readable in flashes.
The Cinematic Intro
The 4-second intro layers Stranger Things alien crickets (the same recipe as plan9_emerge and arp_cathedral so the Napkin Films universe sounds connected) with a fat orchestral horn stab, a 3.5-second 55Hz sub-bass swell building into the BOOM, a wah-wah filter sweep, a crystal-arp ascending, and a hard drop into the song.
Final Spec
| Detail | Value |
|---|---|
| Duration | 175.7 s |
| Mix | -13.9 LUFS, -2.3 dBFS, LRA 6.0 |
| Song | 88 BPM, D minor, 60 bars, 15 channels |
| Voice events | 110+ in the unified vocal stem |
| Samples | 5 WhatsApp memos chopped and looped |
| Voice cast | 3 custom ElevenLabs voices plus 1 back-up vocalist |
| Visuals | PIL stick figures, EQ bars, lasers, confetti, crowd silhouettes |
| Manifesto | 2 streams scrolling for the full 175 seconds |
| Versions | v1 through v10 (10 archived passes) |
How It Fits the Bigger Picture
This is the eighth Napkin Films production. The stack stays the same: PIL for animation, ChipForge for synthesis, ElevenLabs for voice, FFmpeg for assembly, Claude in agent mode for direction. What changes with each film is how far the system gets pushed.
Agent Mode pushed the sample-chopping pipeline harder than any prior production. Real friend voice memos as DJ loops, autotuned and bar-aligned, carrying a festival-trap beat: that is a different class of integration than pure TTS dialogue. The film also introduced the rotating dual-stream manifesto overlay as a visual language that can reappear in future productions.
The characters OG Glenda Bunny and OG Bobby Johnson are now fully documented with canonical voice IDs and visual specs. They will return.
For the full development workflow behind how these films get made in a single agent session, Four Films From Code covers the stack in detail. If you want to understand how agent mode works as a daily development practice, Plan 9 Rap Battle shows the prior production in the same universe. And for the broader context of agentic development as a creative methodology, Agentic Development is the full essay.
The film is released under CC BY 4.0. Remix, repost, drop it into your own thing: credit "Napkin Films / Organic Arts LLC" and link the license. The five WhatsApp samples are friend-and-creator voice memos used with consent. Everything else is procedurally generated from ChipForge and NapkinFilms code.
Watch: https://youtu.be/bMBzJusG78w