audiocraft-audio-generation

Generate music from text prompts using MusicGen by providing a descriptive text string and configuring generation parameters like duration, top_k, temperature, and cfg_coef to control the audio output's characteristics and length.

Create sound effects from text for product prototyping using AudioGen models to produce ambient sounds and effects, enabling quick media prototyping and research tasks directly from natural language descriptions.

You need PyTorch to run AudioCraft audio generation models, along with Python and the specific MusicGen, AudioGen, or EnCodec models, to handle text-to-sound and text-to-music workflows effectively.

Melody-conditioned music generation in AudioCraft works by using MusicGen variants that accept both text prompts and an existing melody input, allowing you to guide the musical structure while generating new audio from the text description.

Text-to-sound generation quality and duration are controlled by parameters including duration for length, top_k for sampling choices, temperature for randomness, cfg_coef for classifier-free guidance, and specific sample rates for audio fidelity.