Bridging the Gap: Making AI Tools More Accessible

"Exploring the need for intuitive open-source AI interfaces that empower artists and creators."

Hello to all my readers,

The sheer volume of AI software releases every week is phenomenal, and today I want to share my personal struggle as an aging woman trying to stay afloat in this ever-evolving landscape. As you know, I am an artist working with AI tools, striving not just to experiment with them but to make meaningful use of their capabilities.

With the rise of Chinese AI startups, the trend toward open-source AI software is accelerating, requiring users to navigate software platforms and download these tools independently. Visual and sound AI software is becoming increasingly accessible to everyday creators, but for people like me, we often have to wait until such software becomes more widespread and user-friendly. To make things more challenging, I struggle with workflows like ComfyUI, which rely on nodes and customization.

(My beloved computer is a 2018 iPad Pro, and while I do a lot with it, I am unable to run certain software—on top of already having limited knowledge of open-source systems.) We need programmers and startups to develop intuitive interfaces for open-source software so that more people can benefit from these advancements.

One of the pieces of software I would love to see on the market is:

”NotaGen is an advanced AI model designed to generate high-quality classical sheet music. Developed through a collaboration between institutions like the Central Conservatory of Music, Beihang University, and Tsinghua University, it employs training paradigms similar to large language models, including pre-training, fine-tuning, and reinforcement learning. The model was pre-trained on a vast dataset of over 1.6 million musical pieces and fine-tuned on approximately 9,000 high-quality classical compositions, allowing it to produce stylistically authentic compositions across various classical genres.”

AIVA is also on my radar, along with Suno and Udio.

For now, I want to extend my heartfelt congratulations to Riffusion, an AI software that I find truly innovative in its implementation strategy. Riffusion is free and trains its model using our renderings, yet it is already sophisticated enough to capture our creative flair. Additionally, it offers users credits—the more we engage with it, the more bonuses we receive. I see this as progressive thinking, and I wish the Riffusion team a successful and prosperous future.

This week, I created several new Riffusion playlists, and I warmly invite you to visit my YouTube channel: Youtube.com/@titacostasound

As a little teaser, here are three songs:

One in Portuguese Fado:

Another in English:

Or how about an instrumental piece?

One particularly interesting aspect of Riffusion is its ability to generate European Portuguese songs with a remarkably good accent. We are so used to the default Portuguese being Brazilian Portuguese that this comes as a pleasant surprise. While Riffusion doesn't always follow the prompt—often generating lyrics in Brazilian Portuguese even when I specify European Portuguese—when it does, the results are truly impressive.

You can find more playlists featuring jazz, Celtic, and ambient music on @titacostasound on YouTube. Enjoy!

See you next week

Tita Costa