HAKORADev/VODER

What it solves

VODER brings professional-grade voice processing to your own machine, completely free and offline. It removes the need for multiple separate tools or cloud subscriptions by packaging 8 core audio modes – speech-to-text, text-to-speech, voice conversion, music generation, speech enhancement, sound effects, vocal separation, and speaker diarization – under one command-line interface. It also tackles the pain of working with messy, multi-speaker recordings, noisy audio, or language barriers through features like speaker separation, dubbing, and any-to-any translation.

How it works

VODER orchestrates a stack of open-source AI models behind a single CLI. You call one command like python voder.py tts or stt, and it routes to the appropriate model: Whisper for speech recognition, Qwen3-TTS and Fish Audio S2-Pro for voice synthesis and cloning, Seed-VC for voice conversion, ACE-Step for music generation, TangoFlux for sound effects, UniSE for enhancement, and pyannote for diarization. It also has a smart input pipeline that accepts URLs from YouTube, TikTok, and other platforms, verifies they point to video, and downloads automatically. Beyond the 8 modes, three task layers add utility: train saves reusable voice clones, quest provides lightweight side-quests like URL download and audio manipulation, and chains lets you wire multiple tasks into custom pipelines (e.g., generate music, isolate vocals, train a voice, dub a video). A separate Project Eva DLC plugs in text-to-image, text-to-video, 3D scene generation, and a local chat (VADAR) using models like Flux 2 Dev and MiniMax H3, each running in its own isolated environment.

Who it's for

VODER is aimed at content creators, podcasters, video editors, musicians, voice actors, and anyone needing professional audio processing without cloud costs or privacy trade-offs. It suits both CLI users and those who prefer a GUI (available). Because all modes run on CPU, it's accessible even without a dedicated GPU, though a GPU speeds things up.

Highlights

• 8 main processing modes in one tool: TTS, STS (voice conversion), TTM (music), STT (transcription), SE (enhancement), SFX (sound effects), SVS (vocal separation), SS (speaker separation).
• Multi-speaker dialogue system with script directives for per-line timing, volume, and duration, plus embedded sound effects and automatic background music.
• Voice design and cloning: describe a voice in plain English or clone from a reference clip, mix both in the same dialogue.
• Language conversion and dubbing: translate speech while preserving the original speaker's voice, with per-segment timing alignment for videos.
• Smart input pipeline: paste a URL from YouTube, TikTok, Bilibili, Snapchat, Instagram, Facebook, or X/Twitter; upload an image for OCR.
• Task layers: train saves reusable voice clones as .tts/.ttse files; quest offers media manipulation utilities; chains compose custom pipelines.
• Project Eva DLC expands to text-to-image, text-to-video (with audio), video editing, 3D world generation, and an uncensored local chat (VADAR) – all via the same CLI patterns.
• Works offline, runs on CPU or GPU, and requires only FFmpeg as an external dependency.