bakrianoo/mazinger

What it solves

Mazinger Dubber provides an end-to-end automated pipeline for dubbing videos. It eliminates the need to manually handle separate tools for downloading, transcribing, translating, and synthesizing speech, allowing users to produce a fully dubbed audio or video file from a single command or via a web interface.

How it works

The system chains ten resumable stages: Download, Transcribe, Thumbnails, Describe, Review, Translate, Re-segment, Speak, Assemble, and Subtitle. It caches the output of every stage, meaning if a process is interrupted, it can resume from where it stopped. The pipeline integrates several AI engines:

• Transcription: Uses CohereX (default in Studio), Faster Whisper, or Deepgram Nova 3.
• Translation: Leverages LLMs (including local options via Ollama).
• Speech Synthesis: Employs voice-cloning TTS engines like Qwen3-TTS and OmniVoice.
• Audio Processing: Uses Demucs for background-audio separation.

Who it’s for

It is designed for content creators and developers who want to translate video content into different languages while maintaining the original speaker's voice or using predefined voice themes, without needing to manage complex individual AI toolchains.

Highlights

• Automatic Voice Cloning: Can automatically extract a 20–60 second sample from the source video to clone the original speaker's voice.
• Flexible Transcription: Supports local transcription (CohereX, Faster Whisper) and cloud-based options (Deepgram).
• Comprehensive Output: Capable of producing dubbed videos with burned-in, styled subtitles including RTL support for languages like Arabic.
• Web UI: Includes "Mazinger Studio," a Gradio-based interface for users who prefer not to use the command line.
• Modular Execution: Each stage of the dubbing process can be run as a standalone command or as part of the full pipeline.