Spaces:

---

Reza2kn

/

representation-chizzler

like 2

Running on Zero

I've been working on Representation Chizzler, a tool for automatic cleaning of speech content in ASR / TTS datasets, and while my focus is on Farsi (Persian 🇮🇷) the tool is language-agnostic and applies to all human speech.

I finally got it working so it chunks + streams audio straight FROM & TO datasets hosted on the hub, and cleaned the dev split (N=365) of MohammadGholizadeh/fleurs-farsi and pushed it here: Reza2kn/fleurs-farsi-representation-chizzler

I was working on the test set (N=871) that i ran into ZeroGPU Quota rate limits, although i am a Pro user. I would really appreciate some sponsored GPU hours so i can run the rest of this, and other audio datasets through it and have 'em cleaned and ready for training! 🙏