Artsy: Audio to Art

Overview of My Submission

 
Artsy is a fun app that allows you to record audio and use the transcribed text to generate art.

👉 Demo me

 

Submission Category:

• Wacky Wildcards

 

Link to Code on GitHub

arndom / artsy

Artsy is a fun app that allows you to record audio and use the transcribed data to generate art.

---

About Artsy 🎨

Artsy is a fun app that allows you to record audio and use the transcribed data to generate art...

Well, that was the plan but some issues occurred down the line during the art generation. The idea was to use public ML models to drive this process which would have yielded some beautiful results, i.e matte painting of a whale in the sea, generated @pixray

The issue lies in the implementation, I attempted to deploy a container instance of the above model on GCP but realized I was way over my head as the GPU costs were very substantial. So, after the setback, I moved to an API but that had its issues with CORS and getting in touch with their team in the remaining time wasn't an option so I decided to run with what works and present it as a 'feature'.

Anyway, if a…

View on GitHub

 

Additional Resources / Info

In actuality this is closer to a simple submission than a finished product, the main reason being the art generation.
I would have loved to have something polished but certain issues came about:

• Initially, I had intended to deploy a GPU-enabled container instance of the model used by Pixray but that was too expensive as my free tier wouldn’t cover the costs on GCP.

• The fallback, HotpotAI had its issues, particularly CORS on the API URL and the URL for the generated image hosted on AWS. I managed to solve those via proxies in development but couldn’t find a way in production.

In hindsight, if knew I would submit in this state, I would have written this on the 1st 😉.

Anyway, I hope you find this interesting, have a wonderful day ✌.