convez.ai
SkillAutomaçãoAvançado
betweentwomidnights

Gary4Claude – Skill de Produção Musical com IA

betweentwomidnights/gary4claude

Rede de APIs de geração e transformação de áudio para produção musical assistida

Ver no GitHub
InicialPluginScore · de 0 a 100

Skill que integra múltiplas APIs de geração e transformação de áudio para agentes Claude. Oferece acesso a síntese de áudio a partir de texto, continuação de tracks, transfer de estilo, injeção de LoRAs e ferramentas utilitárias para manipulação de WAV.

Instalação

$npx convez-ai betweentwomidnights/gary4claude

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
2
Forks
0
Acompanhando
0
Saúde
28%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Gerar tracks musicais completos a partir de descrições em texto
  • Estender e continuar composições existentes com consistência estilística
  • Aplicar estilos de áudio e adaptadores LoRA a gravações
  • Automatizar pipeline de produção musical com encadeamento de gerações

Veredito

Skill experimental mas funcional para automação de produção musical. Requer familiaridade com workflows de geração de áudio e APIs externas; bem documentado com exemplos práticos de encadeamento de operações.

README

A tradução para o português deste README ainda está na fila.

gary4claude

AI music production skill for Claude Code agents. Gives your agent access to a network of music generation APIs — text-to-audio, MusicGen continuation, structured synthesis, ACE-Step cover/continuation, LoRA style injection, and audio transforms.

The Cast

NameWhat it does
JerryText-to-audio generation with BPM-aware looping (stable-audio + finetunes)
SA3Stable Audio 3 generation, loops, continuation, transform, and blendable LoRAs
GaryMusicGen continuation — chain generations to build long tracks
FoundationStructured synth/sample generation with randomizable presets
CareyACE-Step cover/continuation with LoRA captions; optional stem generation
TerryAudio style transfer — transform audio with a text prompt (MelodyFlow)

Install

Copy the skill files into your project's .claude/skills/ directory:

# From your project root
mkdir -p .claude/skills/gary4claude
curl -sL https://github.com/betweentwomidnights/gary4claude/archive/main.tar.gz | tar xz --strip-components=1 -C .claude/skills/gary4claude

Or clone directly:

git clone https://github.com/betweentwomidnights/gary4claude .claude/skills/gary4claude

Usage

In a Claude Code session:

Use the gary4claude skill to make me a 2-minute DnB track with AI vocals.

Or get specific:

Use the gary4claude skill to make a 2-minute 174bpm DnB track in A minor,
then use Carey's billie LoRA adapter with cover/complete.

Agents should build songs sequentially: make one request, wait for the returned audio, save it, then decide the next step. Do not launch parallel generations unless the user explicitly asks for alternate takes.

Skill Files

  • gary4claude.md — Top-level overview, routing, workflows, audio flow between services
  • jerry.md — stable-audio: text-to-audio, loops, finetune prompts
  • sa3.md — Stable Audio 3: loops, continuation, transform, LoRA blending
  • gary.md — MusicGen: continuation, retry, transform, model switching
  • foundation.md — Foundation-1: randomize + generate, BPM handling
  • carey.md — ACE-Step: cover, complete continuation, LoRAs, captions, optional lego stems
  • terry.md — MelodyFlow: audio transforms via Gary

Utility Scripts

The scripts/ directory contains stdlib-only WAV helpers for agents that should not need to write their own audio glue code:

  • scripts/mix_wavs.py — mix compatible WAV loops/stems
  • scripts/tile_wav.py — repeat a loop to a target bar count, duration, or reference WAV length
  • scripts/trim_wav.py — crop by seconds or bars
  • scripts/wav_info.py — inspect duration, sample rate, channels, and peak
  • scripts/gary_preflight.py — verify a WAV is long enough for Gary prompt_duration and get a tiling recommendation
  • scripts/base64_wav.py — encode/decode WAV files for API payloads

Example:

python scripts/gary_preflight.py seed_4bar.wav --prompt-duration 6
python scripts/tile_wav.py drums_4bar.wav drums_8bar.wav --bars 8 --bpm 174
python scripts/mix_wavs.py drums_8bar.wav synth_8bar.wav -o seed.wav --gain 0.8 --gain 0.7 --normalize
python scripts/tile_wav.py seed.wav seed_2min.wav --seconds 120

API Base URL

All services are hosted at https://g4l.thecollabagepatch.com

License

MIT

Ver o original no GitHub

Bom para

Relacionados