Skip to content

Claude/gallant satoshi#2

Merged
leoneperdigao merged 7 commits into
mainfrom
claude/gallant-satoshi
Feb 16, 2026
Merged

Claude/gallant satoshi#2
leoneperdigao merged 7 commits into
mainfrom
claude/gallant-satoshi

Conversation

@leoneperdigao

Copy link
Copy Markdown
Owner

No description provided.

leoneperdigao and others added 7 commits February 16, 2026 11:38
- Add BarkEngine and ChatterboxEngine as new vocal synthesis backends,
  supporting neural TTS with verse-level synthesis for natural prosody
- Enhance VocalProcessor with SSML support for Edge TTS, content-aware
  pitch/rate variation, and improved humanization (per-sentence jitter)
- Add comprehensive tests for new engines and vocal processor features

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add llm_generator.py with Ollama-powered lyric generation (falls
  back to template generator when Ollama is unavailable)
- Add prompts.py with mood-specific prompt templates for the LLM
- Wire lyrics engine selection into pipeline and config defaults
- Update config defaults for levels, swing, and velocity variation
- Update pipeline to pass sentence texts for content-aware processing

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Rewrite drum patterns with hip-hop grid (16th-note resolution),
  ghost notes, and multiple variations per style for organic feel
- Add 'triphop' pattern style with 16th-note hi-hats (~16 events/bar)
  matching Vegyn/Headache production density
- Enhance arranger with per-bar pattern rotation, swing model on
  off-16th positions, per-instrument velocity variation and jitter
- Make verse sections use config-driven drum style instead of hardcoded
- Rework drum synthesis: warmer kick, tighter hi-hat, normalized clap
- Fix tests for updated pattern API (list[dict] return type, durations)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add analyzer.py: extract sonic fingerprint from any audio file (BPM,
  6-band frequency balance, spectral shape, dynamics, stereo, MFCCs)
- Add comparator.py: side-by-side comparison with per-metric severity
  scores and weighted overall similarity (0-100%)
- Add parameter_mapper.py: translate analysis deltas into actionable
  config parameter suggestions (e.g. "raise kick +3dB")
- Add report.py: formatted terminal output matching quality.py style
- Rewrite CLI with subparser support: 'mood-machine analyze <ref.wav>
  --compare <gen.wav> --suggest' while keeping backward compatibility
- Add 39 tests using synthetic audio (sine, clicks, bandpass noise)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Calibrate synthesis and mixing against 'The Beginning of the End'
(Headache, 2023) using the new analysis module. Similarity improved
from 48% to 72%.

Synthesis:
- Add upper harmonics (4th, 5th partials) and octave-doubled layer
  to pad synth, pushing energy into 500Hz-2kHz mid-range band
- Low-mid gap closed by 7.7dB, mid gap closed by 7.3dB

Mixing:
- Add Haas effect (1-2.5ms L/R delay) on hi-hat, clap, snare for
  stereo image from mono drum stems
- Widen drum panning (hi-hat 0.12→0.20, clap counter-panned -0.10)
- Add configurable stereo_widen_amount (was hardcoded at 0.6)
- Add subtle master stereo widening in the bus chain

Config:
- Tune melancholic.yaml: BPM 75→70, triphop drums, raised pad/kick/
  hihat/bass levels, pad_filter_cutoff 1400→1800, velocity 0.20→0.38
- Add spotdl and chatterbox as optional dependencies
- Add reference_tracks/ to .gitignore

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@leoneperdigao leoneperdigao merged commit 879fe2c into main Feb 16, 2026
5 checks passed
@leoneperdigao leoneperdigao deleted the claude/gallant-satoshi branch February 16, 2026 15:06
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant