title	Octave
description	A free, open, AI-native music studio for hobby singers and content creators. Solo artist in, full song out.
tags	daw audio ai mcp midi music-generation free open-source
status	Phase 0 — pre-foundation (vision phase)

Octave

Sing the song. Octave plays the orchestra.

Octave is a free, open, AI-native music studio. It is being designed for the singers, content creators, bedroom artists, indie filmmakers, and home album-makers who can't afford a real one. Plug in a cheap mic, sing your song, prompt for a backing arrangement, let the AI mix and master it — and ship anything from a 15-second Reels hook to a 4-minute single to a 12-minute short-film score, all from the same project.

Important

Octave is currently in Phase 0 — pre-foundation. There is no code yet, only the vision. The full plan lives in PLAN.md.

Who it's for

Not the established pro. Octave is being built for:

• The bedroom singer with a $30 mic and a real voice
• The Instagram / TikTok / Shorts creator who needs original audio fast
• The home album-maker who wants to ship a real EP to Spotify
• The indie filmmaker scoring a short film without a composer's budget
• The podcaster who needs an intro / outro / bed in the show's voice
• The pianist who dreams of a string quartet behind them
• The songwriter who hums melodies but can't notate or arrange
• The non-engineer who can sing but doesn't know what compression is

Important

Octave is not a "short-form" tool. From a 15-second Reels hook to a 4-minute single to a 12-minute short-film score — Octave produces studio-quality output for any listening context. Same engine, same quality bar, every time.

Pros are welcome — but they are not the design center. Yet.

What it will do

flowchart LR
  MIC[🎤 Sing] --> REC[Record]
  REC --> PROMPT[Prompt: 'warm lofi piano, soft drums, C minor']
  PROMPT --> AI[AI generates backing]
  AI --> REFINE[Refine: 'add violin in chorus']
  REFINE --> MIXMA[AI mix + master]
  MIXMA --> EXPORT[Spotify master + Reels cut + film cues]

Loading

End to end, under 30 minutes. No studio. No band. No bill.

The mission in one line

Let one person, with one mic and one laptop, make the song they hear in their head — at a quality you'd hear on the radio — for free.

Vision documents

• PLAN.md — the full vision, goals, feature pillars, roadmap, and success metrics
• docs/modules/ — per-module technical design plans (one file per bounded module)

How Octave is built

Important

Octave is built one small, fully-understood module at a time. No code ships without an approved plan.

Every module — record audio, polyphonic pitch correction, one-tap Reels export, Focusrite hot-plug — gets a deep technical plan in docs/modules/ before any code is written, covering:

• Hardware (the actual device, its limits, its electrical / acoustic behavior)
• Drivers / kernel APIs (ALSA / JACK / PipeWire / Core Audio / WASAPI / ASIO)
• OS / cross-platform abstraction
• Engine layer (buffers, threading, lock-free constructs)
• DSP / algorithm layer (named algorithms, math, papers cited)
• Data and wire formats with byte layouts
• Performance budgets as concrete numbers
• Real-time discipline (RT vs non-RT boundary)
• Failure modes and recovery
• Public API surface (typed, documented)
• MCP exposure (every API gets a tool)
• UI surface (Simple Mode + Studio Mode)
• Test strategy and acceptance criteria

The plan template is enforced by the project-scoped module-plan Claude Code skill at .claude/skills/module-plan/SKILL.md. It inherits the doc-to-dashboard markdown conventions for output. Most of our development time is spent understanding before implementing.

Status

Phase	Description	Status
0	Foundation — audio engine, hardware I/O, APIs	🟡 Pre-foundation, designing
1	The Recorder (vocal-first)	⏳ Planned
2	The Editor + AI cleanup	⏳ Planned
3	The Tuning Room	⏳ Planned
4	Prompt-to-Music (AI generation MVP)	⏳ Planned
5	The MCP Layer	⏳ Planned
6	AI Mix and Master (one-click)	⏳ Planned
7+	Manual depth, MIDI, Virtual Session Musicians	⏳ Planned

See PLAN.md for the full roadmap.

Principles

• Free and public, forever. No freemium. No paywalls. No locked features.
• AI-native. Generative music, AI virtual session musicians, AI mix and master are core — not bolt-ons.
• API-first. Every UI feature ships with a typed API. The UI is just the first client.
• Pro audio quality. 32-bit float internal, sample-accurate automation, plugin delay compensation. Free does not mean cheap.
• Local-first. Runs offline. No required account. No required cloud. Your voice doesn't leave your machine unless you say so.
• Hardware respect. Focusrite day-one. Linux first, then macOS, then Windows.

Contributing

The project is in vision phase. Code contribution starts at Phase 0 (foundation). For now, the most valuable contribution is feedback on the vision — open an issue with thoughts, missing features, or audience perspectives.

License

To be decided. Intent: a permissive open-source license consistent with "free for all, forever."

---

Free for all. That's the dream. — And I will be the first user.