Raven, credited as rzvzn on Discord, trained Kokoro, an open-weight text-to-speech model. The original demo names Raven, and the model card credits the Discord handle. Public evidence supports this specific contribution; it does not establish a legal name or ownership of similarly named accounts.[1][2]
Start with Kokoro →Links
Kokoro model cardContributions
Kokoro generates speech from text with an 82-million-parameter model. Its model card separates rzvzn’s training contribution from the StyleTTS 2 architecture credited to Yinghao Aaron Li and collaborators. The released model supports inspection and local inference; the card documents the particular training-data sources and licenses. These credits establish different contributions to a shared release.[2]
Tools, writing & talks
Tools
Start here
Kokorohuggingface.coAn open-weight text-to-speech model with a hosted demo and inference examples.
Begin with the model card’s audio sample and hosted demo. Compare pronunciation on a short public text; consult the inference example before setting up a local installation.