Skip to content

DAWggies

DAW (Digital Audio Workstation) for AI agents.

Code Video Cerebral Valley

Core trick

Typed agent tools observe an audio timeline and align a sound-effect cue's acoustic peak to a narration syllable's timestamp.

Tags

creative/musicagents

Description

DAW (Digital Audio Workstation) for AI agents. This feature helps Dramatis - AI orchestrator for full-cast audiobooks from raw text - to finely adjust the audiotracks for better immersivity and audio results.
helLf1nGer/Dramatis-DAW★ 00 forksPythonCreated Jun 13, 2026 · last push Jun 13, 2026

README excerpt

# Dramatis DAW — Timeline Surgeon

**Timeline Surgeon** is an *agent-facing* DAW (digital audio workstation) control surface for
AI-produced narrative audio — part of **Dramatis**, an agentic full-cast audiobook studio. Instead
of a human GUI, it gives the Director and sub-agents typed tools to **Observe** an audio timeline,
**Act** on it (align / move / trim / gain sound-effect cues), and **Verify** the result — all
deterministically.

**Headline capability:** align a sound effect's acoustic *peak* to the narration's stressed
syllable — e.g. land the apex of a giant's laugh exactly on the word *"laughed,"* instead of guessing
a timestamp.

## Live demo

`docs/` is a self-contained static page (visual-led): the **before/after timeline** of *"The Cave of
Polyphemus,"* showing the laugh cue slide into alignment, plus a 0:40–0:55 audio comparison. Enable
GitHub Pages (Settings → Pages → deploy from `main`, `/docs`) to serve it as a live URL.

## Two implementations

Built during a hackathon as two independent takes on the same spec:

- **`hellfinger/`** — helLf1nGer's implementation (spec-driven, adversarially reviewed).
- **`gepa/`** — tanny2109's implementation, evolved with a **GEPA** prompt-optimization loop
  (`gepa/optim/`).

Each carries the Timeline Surgeon DAW modules + tests:

| File | Role |
|------|------|
| `dramatis/audio/asset_timing.py` | asset peak / attack / tail analysis (incl. band-limited high-peak) |
| `dramatis/audio/timeline_index.py` | read-model timelin

Team