AuroraNemoia’s Twitter Archive—№ 1,584

@HopeBeyondDitto It's doable, about like 3-4 days of work and a couple hours of training. Rip streams, yeet bgm with demucs, whisper, feed to larger LLM to turn into chatml dataset, feed to unsloth, train mistral-7b qlora, train rvc on like 5 minutes of voice, apply on top of something like xtts.