A self-hosted voice studio — your own ElevenLabs, running on your graphics card.
Turns text into speech in 15 curated voices, clones a voice from a clip as short as six seconds across 17 languages, transcribes audio with word-level timing, and narrates long pieces into a single stitched file. It syncs itself from GitHub, starts the backend, and opens the studio for you.
Voice generation and cloning run on your own graphics card, using models totalling about 2.6 GB. Hosting it would mean renting a GPU server, so it stays on the machine that already has one.