Major Update v2 - SE Mode, SFX, Script Directives

04/08/2026

• Status: Stable, all features work, still developing
• Major Update — Two New Modes, Script Directives, SFX Integration, and Speech Enhancement

Added

• SE (Speech Enhancement) Mode - Audio quality improvement using UniSE model
• SFX (Sound Effects Generation) Mode - Custom sound effects from text using TangoFlux
• Script Directives - Per-line control over timing, volume, and duration
• SFX Character in Dialogue - Embed sound effects directly in dialogue scripts
• Enhanced Dialogue Assembly - Complete rewrite of dialogue generation pipeline
• Cross-use Feature - Mix generated and cloned voices in the same dialogue
• Music Volume Level Control - Fine-grained control over background music volume

Changed

• Expanded Processing Modes - VODER now has 9 processing modes (up from 7)

Enhancement - Music Chunking and Duration Fix

04/03/2026

• Status: Stable, all features work, still developing

Added

• Background Music Chunking for Long Dialogues - Generate multiple music chunks for dialogues longer than 250 seconds

Fixed

• Music Generation Minimum Duration - Enforced 10-second minimum for music generation to match ACE-Step model requirements

Major Update v1 - STT, Diarization, OCR, YouTube

04/02/2026

• Status: Stable, all features work, still developing

Added

• STT (Speech-to-Text) Standalone Mode - Transcribe audio, video, image, or YouTube URL to text
• Speaker Diarization (Pyannote Integration) - Identify and separate speakers in audio
• Image Text Extraction (EasyOCR) - Extract text from images
• YouTube and Video Platform Download (yt-dlp) - Download audio from YouTube, Bilibili, TikTok
• Automatic Voice Clip Extraction - Extract individual speaker voice clips automatically

Changed

• Centralized Model Management System - Complete overhaul of model storage and caching

Major Update - MSTS and Memory Offloading

02/24/2026

• Status: Stable, all features work, still developing

Added

• MSTS (Music-STS) in STS mode - STS now supports musical inputs via the Seed-VC v1 model
• TTS+VC dialogue voice cloning stability - Voice characteristics extracted once per character

Optimized

• Memory offloading after processing - Models explicitly unloaded from memory/VRAM

Major Update - Dialogue Support

02/12/2026

• Status: Stable, all features work, under aggressive testing, still developing

Added

• Full dialogue support in CLI - Both interactive and one-liner modes now support multi-speaker scripts
• Optional background music for dialogue scripts - Available in TTS and TTS+VC modes
• Row-based dialogue editor in GUI - Replaced free-text script box with per-row Character/Dialogue fields

Fixed

• Memory optimisation for TTM+VC

Seed-VC v2 Fix

02/10/2026

• Status: Stable, all features work, under aggressive testing, still developing

Fixed

• Seed-VC v2 unmatched tensor error which caused both STS and TTM+VC to fail. Now STS works perfectly; TTM+VC will receive further optimisations.

Initial Release - Unstable Development Build

02/09/2026

• Status: unstable, untested, under development

Initial Release - Unstable Development Build

First public release of VODER.