Voice Studio v0.7.2

Full Changelog: v0.7.1...v0.7.2

Voice Studio v0.7.1

What's Changed

• feat: Argos Translate — offline in-process translator (LibreTranslate engine) by @wpacademy in #42
• feat(dub): separate controls from transcript + popular language chips by @wpacademy in #43

Full Changelog: v0.6.1...v0.7.1

Voice Studio v0.6.1

What's Changed

• feat: Cross-language dubbing + offline translation (M2M-100, MADLAD-400) by @wpacademy in #40
• fix: translation lineup (M2M-100 1.2B) + download cancel + trim wasted GB by @wpacademy in #41

Full Changelog: v0.5.2...v0.6.1

Voice Studio v0.5.2

Full Changelog: v0.5.1...v0.5.2

Voice Studio v0.5.1

What's Changed

• feat(dub): Design/Auto voice modes + Qwen style in Dub mode by @wpacademy in #39

Full Changelog: v0.5.0...v0.5.1

Voice Studio v0.5.0

What's Changed

• fix(logging): mute phonemizer's cosmetic 'words count mismatch' warnings by @wpacademy in #37
• feat: Voice-to-voice dubbing (Dub mode) by @wpacademy in #38

Full Changelog: v0.4.0...v0.5.0

Voice Studio v0.4.0

What's Changed

• docs(readme): clarify the 5000-char cap is app-wide by @wpacademy in #34
• docs(readme): add Qwen3-TTS + Whisper/Transcribe, fix engine count by @wpacademy in #35
• feat: add Kitten TTS Mini — ultra-light ONNX engine for low-end hardware by @wpacademy in #36

Full Changelog: v0.3.3...v0.4.0

Voice Studio v0.3.3

What's Changed

• perf: provision venvs with uv to cut disk ~27GB → ~13GB by @wpacademy in #11
• feat: hardware status bar (CPU/RAM/VRAM/disk/cache) by @wpacademy in #10
• feat: music generation tab (ACE-Step 1.5) by @wpacademy in #12
• feat: music metadata controls + batch generation + WAV/FLAC downloads by @wpacademy in #14
• feat: music-mode left controls + sample presets by @wpacademy in #13
• feat: music AI Inspiration + Thinking (ACE-Step 0.6B LM) by @wpacademy in #15
• feat(music): Cover / Remix + Repaint (audio-to-audio) by @wpacademy in #16
• feat(music): 2B base model + Extract / Lego / Complete by @wpacademy in #17
• fix(music): text2music breaks into noise — default to LM (thinking) by @wpacademy in #18
• refactor(music): remove ACE-Step, keep Music as an engine-agnostic plug-in point by @wpacademy in #19
• feat(music): MusicGen-small engine (in-process, 2.4 GB, CC-BY-NC-4.0) by @wpacademy in #20
• Remove MusicGen and the music stack entirely by @wpacademy in #21
• ASR (PR A): Whisper large-v3-turbo backend + GpuGate by @wpacademy in #22
• ASR (PR B): Transcribe mode UI by @wpacademy in #23
• ASR (PR C): reference_transcript auto-fill + subtitles by @wpacademy in #24
• fix(status-bar): device-wide VRAM + equal-width chips by @wpacademy in #25
• feat: RTL-aware direction for text inputs by @wpacademy in #26
• fix(setup): model picker reads the real catalog by @wpacademy in #27
• fix(setup): force UTF-8 console so fresh Windows setup doesn't crash by @wpacademy in #28
• fix(tts): cache multi-line takes + Download button beside Play by @wpacademy in #29
• chore: bump version to 0.3.0 by @wpacademy in #30
• fix(ui): Whisper card in Engines popup + restore Transcribe sidebar chrome by @wpacademy in #31
• fix(ui): pin Appearance + Backend to bottom of Transcribe sidebar by @wpacademy in #32
• fix(ui): Speech-to-text section scrolls with the engine cards by @wpacademy in #33

Full Changelog: v0.2.0...v0.3.3

Voice Studio v0.2.0

Voice Studio by MSR — v0.2.0

The first tagged release of Voice Studio, a fully-offline, local web UI for
multiple open-source text-to-speech engines. Everything runs on your own
machine — no cloud, no API keys, no audio leaving your computer.

Engines

Six TTS engines, loaded one at a time to keep memory low:

• VibeVoice-1.5B — expressive multi-speaker synthesis
• Kokoro-82M — fast, lightweight, multilingual
• Chatterbox Multilingual V3 — 23 languages with voice cloning
• OmniVoice — clone / design / auto voice modes
• VoxCPM2 — 2B model, 48 kHz, 30 languages, ultimate cloning
• Qwen3-TTS CustomVoice — 9 premium built-in voices, 10 languages, free-text style control

Highlights

• Podcast and Text-to-Voice project modes — multi-segment editor or a single textarea, on any engine
• Voice library with cloning, built-in voices, and per-voice metadata
• Three-column ElevenLabs-style UI with light/dark themes
• One engine at a time, isolated venvs for engines with conflicting dependencies (Chatterbox, OmniVoice, VoxCPM, Qwen)
• In-app model downloads with live progress, and in-app engine installs for isolated engines
• Versioning + auto-update — checks GitHub Releases and applies updates in place on git checkouts
• Synthesis cache with thread-safe, atomic writes

Getting started

python studio.py setup    # one-time: venv, deps, CUDA auto-detect, model picker
python studio.py start    # run backend + frontend

Built by Mian Shahzad Raza (MSR) — https://msrbuilds.com/