You Can Use AI To Generate Music: Stable Audio Open
By Genpakt
Summary
## Key takeaways - **Stable Audio Open 1.0 Announced**: Stable Audio Open version 1.0 just got announced and published with open source weights for generating different audio using text prompts only. [00:00], [00:39] - **Install Packages in Colab Pro**: Install the listed packages using Google Colab Pro Plus with an L4 GPU, then restart the environment and install Hugging Face Hub. [00:50], [01:01] - **Login for Model Access**: Log in with Hugging Face token because the model requires access grant, and store the token in environment secrets for security. [01:12], [01:35] - **Set Text Prompt and Timing**: Set up text prompt like 'make a two pack style bob beat' with timing from zero to second 30. [01:55], [02:07] - **Run Diffusion to Generate Audio**: Define diffusion parameters including steps, conditioning, sample size, execute with the model, rearrange audio batch, and save as WAV file. [02:17], [02:36] - **Download and Play Generated Beat**: The generated audio saves to the local machine for download and playback, producing a two pack hip hop beat. [02:45], [03:04]
Topics Covered
- Stability Audio Open 1.0 Now Open Source
- Text Prompts Generate Specific Beats
- Run on Free Colab L4 GPU
Full Transcript
stable audio open version 1.0 just got announced and got published these are the different audio that I've been able to generate here's another one
[Music] and there's many other examples they can use to generate various beats and music
using text prompts only and they just announced and published The Open Source weights and today I'm going to show you guys how to use them so first thing you have to do is
install all the different packages that I've listed over here I am using Google collab Pro Plus with a L4 GPU if in if in case people were Wonder
ring now it is asking us to restart the environment we've done that all we have to do is we have to install hug and face Hub which we have done over here now we just have to make
sure we log in and this is needed because we need to access a model that requires access Grant we just paste our token in and then we're good to go by the way I make
similar videos like these on lm's machion learning and other data science tools so please feel free to subscribe now all we do do is import our packages we decide a device we download
the model and we put the model to the device for faster inference now I just want to mention that I've also described the HF token in my environment secrets
so over here it's probably a good idea to do that now once that is done we have to set up our text prompt and timeing conditioning so over here we have a
start seconds from zero to second 30 and and our prompt is make a two pack style bob beat we execute that we Define other
diffusion parameters so we pass in the model the steps conditioning sample size and few other information and we also pass in the device information as well
once that is done we have to rearr rearrange our audio batch and also Define which name we're going to save
our output WB file so this will be two pack hip hop now it has saved the generated audio into our local machine
over here now we just have to download this and play it all right he has downloaded let's
play it
[Music]
all right there you guys have it that is how you generate music beats or any audio format Clips using an open source
model called stability audio open 1.0 all right guys hope you guys found insightful if you guys did please feel free to subscribe I make similar videos like these all right guys have a nice
day bye-bye
Loading video analysis...