Dubsmith

What is Dubsmith?

Dubsmith is an open-source command-line tool for dubbing videos while preserving each speaker’s voice timbre. It detects speech and speakers, translates dialogue into one of ten target languages, generates new audio, and combines it with the video. It can also create subtitles without dubbing. A separate speech-generation mode creates audio from saved, designed, cloned, or preset voices. The models run locally; translation and optional prosody planning use a configured OpenAI-compatible LLM endpoint, which can be local or cloud. A GPU is optional but strongly recommended. The code is Apache-2.0, while model weights have separate licenses. The current version does not include lip sync.

Dubsmith features

• Translates video audio into ten languages while preserving original speaker voices
• Runs locally with model weights stored on your own hardware
• Clones individual speaker voices and maintains original timing
• Generates speech from text using custom, cloned, or preset voices
• Applies emotion, style, and prosody controls to generated audio
• Exports detailed quality reports and timestamped provenance logs
• Supports background noise and room acoustics preservation

Who uses Dubsmith and for what

• Localizing independent films or YouTube videos without hiring voice actors
• Creating multilingual training materials while keeping instructor identity intact
• Generating audiobook samples or narration tracks with consistent custom voices
• Producing subtitle-only translations for quick content review before full dubbing
• Testing different emotional tones for podcast scripts using cloned reference voices

Dubsmith FAQ

Comments (0)

No comments yet. Ask the maker a question or share how you would use it.