AudioCraft — генерация музыки и звука из текста

★ 7.1 · media-processing

audiocraft-audio-generation is a Claude Code skill that enables text-to-music and text-to-audio generation using Meta's AudioCraft library. It integrates three core components: MusicGen for music generation with optional melody conditioning, AudioGen for synthesizing sound effects and environmental audio, and EnCodec as a high-fidelity neural audio codec that converts tokens back to waveforms. Model sizes range from 300M parameters (musicgen-small) to 3.3B (musicgen-large), with dedicated stereo variants and musicgen-style for reference-based generation. The skill supports both the native AudioCraft API and HuggingFace Transformers, making it practical for developers building music generation tools, automated audio production pipelines, or applications that need controllable, description-driven sound synthesis.