Open-source OmniVoice Studio dubs video in 646 languages fully local, taking on ElevenLabs

thisdudelikesAI · x · 2026-11-07

A long X thread introduces OmniVoice Studio / VoiceStudio (GitHub: debpalash/VoiceStudio), an open-source, fully-local ElevenLabs alternative offering real-time dictation, zero-shot voice cloning, and end-to-end cinematic video dubbing in 646 languages.

Key capabilities

Drop in an MP4, pick a language: the pipeline transcribes, translates, generates new speech, and mixes it back — no manual editing
Voice cloning from 3 seconds of audio; design new voices by stacking tags (female, elderly, british, raspy, deep)
Uses demucs to isolate vocals so original music, ambience, and SFX survive the dub — a step most paid tools butcher

Deployment

Three-command setup (clone, cd, bun install && bun dev; needs ffmpeg + Bun); model weights auto-download on first use, zero config
Frontend on localhost:5173, API on port 8000; automatic threading, caching, and VRAM management; runs on Apple Silicon, NVIDIA/AMD GPUs, and CPU-only

vs ElevenLabs (author's framing): $99/month Pro with per-minute dubbing vs free and unlimited; cloud-only vs files never leaving your machine.

Caveats: the thread claims Apache 2.0, but the README shows FSL-1.1-ALv2 and GitHub lists AGPL-3.0; star counts also conflict across screenshots (1.6k vs 103k). Verify against the actual repo before trusting the hype.

Original post →

Open-source OmniVoice Studio dubs video in 646 languages fully local, taking on ElevenLabs

More from coding & agent