audiocraft-audio-generation

Text-to-music generation uses MusicGen to create audio content directly from text prompts, supporting melody conditioning and style transfer for video production. Multiple model sizes are available to match your project scale.

AudioGen generates sound effects from text prompts for game audio and interactive experiences. It works alongside MusicGen and EnCodec to deliver high-fidelity neural audio codec outputs.

Melody conditioning is a text-to-music generation feature that allows reference-based style transfer. It enables you to guide the audio creation process by providing a melody reference alongside your text prompt.

Stereo support is included in the audio generation capabilities. The system uses EnCodec for high-fidelity neural audio encoding across its small to large model sizes.

Install the audiocraft package and load a pretrained MusicGen or AudioGen model to generate audio from a text prompt. The workflow requires a SKILL.md frontmatter with a name and description.

Neural audio codec generation requires manual effort reduction but depends on model size selection from small to large. Sound design tasks should consider the appropriate model scale for desired fidelity.