Skip to content

OmniVoice Studio v0.3.7

Latest

Choose a tag to compare

@github-actions github-actions released this 20 Jun 09:32
8e6344e

Stable release built from main. A stabilization pass that clears the
wave of issues reported on the 0.3.6 line — across voice design, dubbing,
transcription, install, and the Linux/web UI — and lands two more opt-in
cloning engines. The throughline: non-English correctness and cross-platform
playback
— cloned and designed voices now hold their language end-to-end, and
audio plays inline in Linux/Android browsers, not just macOS.

🌍 Non-English correctness

  • 🗣️ Voices no longer drift to English / the wrong language — profile + request
    language is threaded through both the single-shot and longform paths, and text
    is NFC-normalized before duration estimation (fail-before/pass-after tested)
    by @debpalash in #565

🔊 Cross-platform audio & UI

  • ▶️ Audio playback on Linux/Android browsers — IANA-canonical audio/wav /
    audio/flac MIME + gesture-driven AudioContext unlock, so the play button
    works outside macOS by @paoloantinori in #511
  • 🐧 Generate / Settings / Clone buttons are pressable on Linux — the shell now
    fills the window on WebKitGTK instead of leaving a black band that pushed the
    CTAs off-screen by @debpalash in #558, #519
  • 🪟 Settings tabs fill the panel instead of a stunted box in a void, and the
    Appearance strings now route through i18n by @debpalash in #507
  • 🎬 Windows: renders play in-app, not in an uncloseable black window
    by @debpalash in #538

✨ Engines & voices

  • 🧩 Two opt-in heavyweight TTS engines — MOSS-TTS-v1.5 (8B) + dots.tts (2B)
    zero-shot cloning in isolated subprocess venvs (no change to the default
    install or its lockfile) by @debpalash in #531
  • 🎭 "Save design as profile" no longer poisons the profile with
    [object Object] (which then 400'd every generation); a migration heals
    already-saved profiles by @debpalash in #556
  • Stories: the global reading-speed slider now applies to preview + stem
    export
    , not just the full render by @debpalash in #509

🛠️ Reliability & install

  • 🩹 Profile / persona / consent endpoints stop 500'ing on
    no such column: consent_audio_path — the schema self-heals missing additive
    columns on startup by @debpalash in #557
  • 📦 Engine "Install" button worksuv pip install targets the running
    interpreter instead of a venv it couldn't find by @debpalash in #559
  • 🎙️ Transcription survives GPUs without efficient float16 — both CTranslate2
    ASR backends fall back float16 → int8 instead of "no segments"
    by @debpalash in #561
  • 📚 Audiobook import no longer 500's (AudiobookPlan.chapter_count) across
    .txt / .md / .epub / .pdf by @debpalash in #544
  • 🌐 Actionable video-download errors + a relocated backend venv self-heals
    (No module named 'encodings') by @debpalash in #562
  • 🔢 Version is single-sourced from frontend/package.json so a build can't
    call itself the wrong version again; the frozen backend now reports its real
    version (no more 0.3.5 fallback), fixing the blank About → Version
    by @debpalash in #503, #501
  • 🔁 Updater Preview channel tracks main again (nightly rebuild) with
    prerelease + Intel-Mac parity guards by @debpalash in #500
  • 💛 Donate goal bar shows real progress ($10 / $200) instead of fabricated
    numbers by @debpalash in #513

💡 Also carries the v0.3.6 "Can't reach the local backend" startup-crash
fixes (pkg_resources, scalar_fastapi, exit-106 broken venv) — updating from
v0.3.5/older resolves them.

🌟 New Contributors

Full Changelog: v0.3.6...v0.3.7

💖 Contributors

@debpalash @paoloantinori


Verify your download

SHA-256 checksums for each platform artifact:

Linux x64 artifacts

487c19c1915e467c35514293a9a8808b17bf1bc35c035cfd7055d72303d6bd8e  OmniVoice Studio_0.3.7_amd64.AppImage
57e04d9f27581eb6ed1f49c739074087641ca4e61b9feb070141fe53e4f0415c  OmniVoice Studio_0.3.7_amd64.AppImage.sig

Windows x64 artifacts

a40154f754719ab780c8b880be444f755d736bfa7a739454c861731817d80eaa *OmniVoice Studio_0.3.7_x64_en-US.msi
8151f634524501616b29aa99a1043d575fd52227d3330a0e3a434975cd3bc451 *OmniVoice Studio_0.3.7_x64_en-US.msi.sig

macOS Intel artifacts

fceb715e0333b51abfd8f6c4e1a1125ec149a09d8ac9c2d840ba941d4fa6d12c  OmniVoice Studio_0.3.7_x64.dmg
6e79c2eae746d2903f37f01e0fc3e6758a77472697b50bddfb316036b9cae08d  OmniVoice Studio.app.tar.gz
92a20548d5fb2ace4347f7d073bcf01b8d620416efb566133a7136d351f2e355  OmniVoice Studio.app.tar.gz.sig