chiptune-cover

name

chiptune-cover

description

Convert well-known melodies, themes, and scores into chiptune arrangements using chiptune-studio. Takes a song reference (title, artist, or melody description), researches key/tempo/notes, makes creative arrangement decisions per scene or section, and outputs a composed WAV via chiptune-studio synthesis. Use when: user asks to "make a chiptune version of [song]", "cover [track] in 8-bit", "arrange [theme] for chiptune", "do a chiptune [song name]", or references any well-known melody that should be rendered through the chiptune-studio toolkit. Also triggers on: "soundtrack this with [song]", "score this video with a chiptune [reference]".

Chiptune Cover

Arrange well-known melodies as chiptune compositions via chiptune-studio.

Prerequisite skill: chiptune-studio (synthesis toolkit). This skill handles the creative arrangement process. Chiptune-bricks handles the sound.

Process

1. RESEARCH — Extract the Musical DNA

Gather these facts about the source material:

• Key/mode (e.g., E Phrygian, C minor, G Mixolydian)
• Tempo (BPM)
• Core melody as note names (e.g., E E D C Bb B)
• Harmonic structure (chord progression, pedal tones, key changes)
• Signature characteristics (what makes it recognisable... the riff shape, a rhythmic pattern, a chromatic move)

Research sources (try in order):

1. Web search for "[song] key tempo notes" or "[song] music theory analysis"
2. Hooktheory TheoryTab if the song is well-known
3. MIDI databases for note-level transcription
4. Your own musical knowledge for standards and classical pieces

Output: A docstring at the top of the composition script documenting what you found.

2. ARRANGE — Map Source to Chiptune Vocabulary

Don't transcribe literally. Translate idiomatically.

Translation principles:

Source sound	Chiptune equivalent
Guitar riff	arp() with sawtooth/square, crush=3-4
Bass line	arp() octave down, square, note_frac=0.5, lowpass(800)
Drum kick	noise(0.06) + lowpass(150) + tight env(a=0.002, d=0.03, s=0.1, r=0.02)
Drum hat	noise(0.04) + highpass(2000) + tight env
Orchestral pad	drone() or pad() with triangle, heavy lowpass
Choir/vocal	Detuned pad() pairs with tremolo, slow env attack
Solo melody	sawtooth/triangle arp with reverb/delay
Power chord	Stack root + fifth as simultaneous arps
Sustain/held note	resolve() with reverb
Distortion	bitcrush depth 2-4

What to preserve vs reinvent:

Preserve	Reinvent
Core melody (the recognisable notes)	Instrumentation (chiptune voices)
Key and mode	Arrangement density (usually sparser)
Rhythmic feel	Production effects (bitcrush replaces distortion)
Signature intervals (what makes it that song)	Dynamics (env/vol curves)
Tempo (unless scene-matching)	Layering (2-4 voices max)

3. SCENE-MATCH — Adapt to Narrative (if scoring video)

When scoring a video with defined scenes:

1. Get scene timecodes from manifest/timecodes file
2. Map source intensity to scene emotional arc
3. Vary arrangement per scene:
   • Quiet: isolated notes, half/third speed, sparse
   • Building: add pedal tone, increase speed, introduce riff
   • Peak: full riff, full speed, doubled voices, sub + noise
   • Reflective: hymnal version (slow, triangle, reverb, pads)
   • Resolution: single resolve(), ghost of melody fading

4. COMPOSE — Write the Script

Use this template structure:

# /// script
# requires-python = ">=3.10"
# dependencies = ["numpy", "scipy", "pydub", "audioop-lts"]
# ///
"""
[Title] — [Source Song] as chiptune [style].
[Key]. [BPM]. [Arrangement approach].
Core: [melody as note names]
Original: [composer]. Chiptune-covering with love.
"""

import sys
import numpy as np
sys.path.insert(0, "~/Repos/symbient-skills/skills/chiptune-studio/scripts")

from bricks import *
from bricks.osc import sine, square, sawtooth, triangle, noise, silence, pad

SR = 22050
DUR = [total_duration]
BPM = [tempo]

c = canvas.make(DUR)

# [Define riff/melody note arrays]
RIFF = ["E3", "E3", "D3", "C3", "Bb2", "B2"]

# ============================================================
# SECTION [N]: [NAME] ([start]-[end]s) — [description]
# ============================================================

# [Place layers with canvas.place()]

# ============================================================
# MASTER
# ============================================================

c = canvas.normalize(c)
c = fade_in(c, 1.5)
c = fade_out(c, 3.5)

canvas.save_wav(c, "[output_path]")

Composition heuristics:

• 2-3 simultaneous layers max. Add only what earns its place.
• lowpass everything that isn't the lead melody
• sub_rumble() under intense sections for physical weight
• Ghost octave doublings (triangle, vol=0.03-0.05) add depth without mud
• sparkle() cascades for transitions or "something appearing"
• End on resolve() at the root note

5. RENDER + MIX

uv run [score_script.py]

Mix levels when combining with voiceover:

• Score alone: normalize to -3dB (default)
• Under voiceover: -5 to -8 dB
• Under dense narration: -8 to -12 dB

Reference

See references/doom-e1m1-example.md for a complete annotated walkthrough of converting Doom E1M1 "At Doom's Gate" into a scene-matched chiptune score.

Anti-Patterns

• Literal MIDI port — a cover is an arrangement, not a transcription
• Too many layers — 2-4 voices. If you need 6, simplify.
• Ignoring the signature — find what makes it that song and protect it
• Flat dynamics — vary density across sections. Dynamic range is the secret weapon.
• Square wave for everything — sawtooth=leads, triangle=hymnal, square=backbone, sine=sub/sparkle
• No percussion — even simple noise kicks give rhythmic grounding
• No ghost doubling — quiet octave-up triangle behind the lead adds depth cheaply