November 24, 2023
Authors
Audiogen Team
Hello everyone!
We're excited to share something we've been working on at Audiogen – our new web app.
It's a small step in the grand scheme of things, but it's a big leap for us in terms of what we've learned and how we hope to grow with our community.
Our journey in creating this app wasn't easy. We faced our fair share of hurdles and had to make some tough decisions along the way. Ultimately, we chose to keep things simple and focus on what really matters – improving the way you control our models.
One key aspect of generative AI is that the models we build can only be as useful as our ability to finely control their outputs. With this release, our aim extends beyond just launching an app; we're also setting the stage for the exciting developments that lie ahead.
We are introducing the concept of conditioning prompt and negative prompt to the Audiogen controls.
These new experimental functionalities enable users to input both sound and text, offering a dual-modality approach to guide the generation process. With these tools, you can now direct the AI not only towards specific outcomes but also steer it away from certain directions, ensuring more precise and tailored results.
We've enhanced the Audiogen model to be able to continue an input sample with a generated sequence.
In our testing, it was particularly interesting to pair the continuation feature with the conditioning prompt and negative prompt.
This is especially true for loops; it's now possible to continue a loop by adding (or removing) instruments to it via conditioning prompt.
This release is but a small step in a long journey to democratize audio creation.
We are planning to soon empower creators and professionals with a state of the art 48khz stereo model, with even more precise and fine grained controls.
Stay tuned for updates, and most importantly:
Thank you all for reading!
.svg){kind=icon}