Open source: cloned Rocky's voice from Project Hail Mary in two days, full pipeline + 2:10 of training audio + trained RVC v2 model

In three linesRocky's voice (Project Hail Mary) cloned in two days via open-source pipeline. Audio extraction (ffmpeg + demucs), transcription (Whisper), diarization (pyannote), then RVC v2 training on 2:10 min audio. Trained .pth model (55MB) and code public. Tested XTTS v2 / YourTTS / RVC v2 / OpenVoice v2.

Summary generated by Claude — human-verified