audiocraft-audio-generation

Load a pretrained model with MusicGen.get_pretrained('facebook/musicgen-small'), set generation parameters like duration and temperature, then call model.generate with a list of text descriptions. Save the returned waveform with torchaudio.save at 32kHz.

MusicGen generates music from text prompts and supports melody conditioning and stereo output at 32kHz. AudioGen generates environmental sounds and sound effects from text at 16kHz. Choose based on whether you need musical content or sound effects.

Yes, MusicGen is available in Transformers as MusicgenForConditionalGeneration with AutoProcessor. Note that parameter names differ: use guidance_scale instead of cfg_coef and max_new_tokens instead of duration.

MusicGen-small needs about 4GB VRAM in FP32 or 2GB in FP16, medium needs about 8GB/4GB, and large needs about 16GB/8GB. Reduce duration or use a smaller model if you hit CUDA out-of-memory errors.

Check the output amplitude and normalize if it is too quiet. For better quality, use a larger model, write more specific prompts describing genre, instruments, and tempo, and tune cfg_coef and temperature.

Use Stable Audio for longer commercial music generation, Bark for text-to-speech combined with music or sound effects, and Riffusion for spectrogram-based generation. AudioCraft fits text-to-music, melody conditioning, and sound effect tasks.