facebookresearch / audiocraft

Audiocraft is a library for audio processing and generation with deep learning. It features the state-of-the-art EnCodec audio compressor / tokenizer, along with MusicGen, a simple and controllable music generation LM with textual and melodic conditioning.
MIT License
20.17k stars 2.01k forks source link

[Question] Generating samples with defined start and proper endings? #412

Open m-brogin-sg opened 5 months ago

m-brogin-sg commented 5 months ago

Hello guys, I am attempting to create a musical intro (e.g. for a YouTube channel) that includes a distinct starting effect, a smooth fading transition (if possible), and most importantly, a proper ending that doesn't get abruptly truncated. The goal is to have the music end gracefully, providing a professional touch to the overall experience.

Starting Effect: Most of the samples that I generated do not have a proper "opening" to the song. I would like the sample to have like an attention-grabbing sound at the beginning.

[OPTIONAL] Fading Transition: The intro music should smoothly fade in at the beginning and fade out at the end to create a polished and seamless transition.

Proper Ending: I can't generate a song that actually ends with an ending and doesn't get truncated away. Instead of an abrupt stop, I want the music to taper off gradually, giving it a natural and pleasing conclusion.

Is there any way or settings to change so the AI can achieve this? Any recommended parameters, techniques, or examples that could help me in creating jingles with these characteristics?

I'm mostly using it from the Gradio web interface, maybe via code you can tweak things better?

Looking forward to knowing if anyone can help me achieve this, thanks :)

mykeehu commented 4 months ago

That's what I would like to do, to have a beginning and an end to the music, and to generate music that can be looped.