Download
| Platform | Download | Signing |
|---|---|---|
| Windows x64 | Installer | ⚠️ Unsigned: SmartScreen shows "Windows protected your PC"; choose More info → Run anyway |
| Windows ARM64 (experimental) | Installer | ⚠️ Unsigned: SmartScreen shows "Windows protected your PC"; choose More info → Run anyway |
| macOS Apple Silicon | DMG | ✅ Signed with Developer ID and notarized by Apple |
| macOS Intel | DMG | ✅ Signed with Developer ID and notarized by Apple |
| Linux x64 | AppImage · deb | — Linux packages are not code-signed |
| Android | APK · AAB | ✅ Signed APK; the AAB is the Play Store bundle |
| iPhone | Unsigned IPA | ⚠️ Unsigned: sign it with your own Apple ID to sideload; not a TestFlight build |
Setup and model downloads recover more reliably after interrupted transfers, low-space retries and transient connection failures. This beta also improves low-VRAM generation, pronunciation, long-recording transcription, plain-text exports in Dubbing and Transcriptions, and Android downloads.
Highlights
- Setup reuses verified wheels and retries transient connection failures; model downloads reject corrupt ranges while preserving valid work
- Low-VRAM GPUs release idle models before generation
- Saved pronunciation entries and inline overrides work across speech, Dub, batch and voice conversion
- Save plain transcripts from Dubbing before or after generation on desktop, or from generated paired-phone results; Transcriptions also exports SRT with genuine cue timings and handles long recordings with duration-scaled limits
- Android downloads work through eligible VPN routes, and installed gated models open their licence review dialog
Testing wanted: small-GPU generation under memory pressure; pronunciation in Dub, batch and voice conversion; long-file transcription and text/SRT export; edited Dubbing transcripts before generation and paired-phone TXT sharing; low-space or restricted-network setup; interrupted model downloads followed by Retry; and Android downloads with a VPN and cellular opt-in.
Known scope: the low-memory Mac Stories performance report #4090, Stories playback report #4136, legacy Dubbing Method Not Allowed report #4127 and Windows native model-loading crash #4142 remain unresolved. Dubbing TXT export answers the separate plain-transcript request in #4127; it does not repair or establish a retest of the archived v0.3.8 Tauri SRT/405 failure. VoxCPM2 installation report #4150 still needs the original failing step and logs. The playback wording does not fix the underlying Stories failure. Download transport pinning applies to accelerated transfers, rather than all Hub metadata or plain Hugging Face fallback traffic.
Upgrading: install over the existing app or use Settings → Updates → Beta. Voices, projects, settings and installed models stay in place. The release workflow adds the download table from actual desktop and phone artifacts, including measured signing results.
Changed
- Removed an unused legacy update-busy helper while retaining active restart protection (#4149) — thanks @themrbeasley!
- Retired an unused legacy update toast while preserving the active Electron update controls (#4141) — thanks @themrbeasley!
Added
- Dubbing exports the selected transcript as plain UTF-8 TXT, including edited text before generating a dub, with native Save As and paired phone sharing. (#4155) — thanks @shivamdhiman999111-source!
Docs
- Track the unresolved VoxCPM2 installation report separately from GPU fallback fixes, reported by vinhson (#4154)
- Document host-managed Tailscale access for Docker without sharing the host daemon socket (#4111) — thanks @MaverickKB!
- Installation tracking separates shipped fixes, pending repairs and unresolved report evidence (#4121)
- Track the unresolved Windows model-loading crash and distinguish merged installer repairs from released fixes (#4146) — thanks @renaldyramadani123-code!
Fixed
- Diagnostic bundles and Electron bug reports redact share PINs and encoded credential query parameters (#4151) — thanks @DeryFerd!
- The reviewed model-install steps now require the same authenticated consumer as the rest of the model manager (#4152) — thanks @DeryFerd!
- macOS publishing checks Gatekeeper even when Windows installers are explicitly unsigned (#4153)
- Malformed model-install status responses keep the interface open and preserve the last valid download state (#4106) — thanks @ftlcustomerservice73-ui!
- Auto setup can retry PyPI failures through the regional mirror and report the actual failed host (#4107) — thanks @mrkeshi!
- Optional analytics no longer delays backend readiness, and consent changes invalidate queued startup work (#4108) — thanks @cf3901646!
- Update the optional Dagger pipeline's OpenTelemetry dependencies to fix excessive CPU use when parsing malformed trace context. (#4109)
- URL imports restore their connection guard after temporary DNS resolver changes (#4110)
- Long recordings use duration-scaled transcription limits across uploads, dubbing, child engines and MCP tools (#4112) — thanks @xabiherdz-svg!
- Stop cancelled setup retries promptly and keep verbose runtime probes from blocking on their diagnostic output. (#4113)
- Saved and inline pronunciation choices apply consistently across speech, dubbing, batch and voice-conversion routes (#4114) — thanks @nntloc!
- Reuse verified setup wheels on low-space retries and handle interrupted size probes without an uncaught error. (#4115)
- Model downloads reject damaged files and retry without reusing corrupt ranges, while cancellation preserves verified work (#4116)
- Reclaim idle GPU models before generation when VRAM or RAM is low (#4117) — thanks @vandyahmad24!
- Crash recurrence reports recognize current Electron version fields and distinguish beta releases (#4120)
- Model deletion over the API now requires administrator rights (#4122) — thanks @DeryFerd!
- SoniTranslate installation and start/stop now require a local administrator, like every other engine installer (#4123) — thanks @DeryFerd!
- Deleting history, rewriting dub scripts and clearing model-licence receipts now require an administrator, as the API docs already promised (#4124) — thanks @DeryFerd!
- Imported project bundles can no longer make ffmpeg fetch remote URLs through a playlist-shaped media file (#4125) — thanks @DeryFerd!
- TTS and event WebSockets now enforce the same loopback/credential rule as dictation, closing a share-PIN bypass (#4126) — thanks @DeryFerd!
- MCP tools preserve remote caller permissions across local and HTTP backends, including failed transport lookups (#4128) — thanks @DeryFerd!
- Docker deployments now require the administrator API key for the runtime log, log stream, recent errors and crash report, as the install docs already said (#4129) — thanks @DeryFerd!
- A deterministic test now fails when a new mutating API route ships without an authorization decision (#4130) — thanks @DeryFerd!
- The LM Studio model probe no longer sends the stored API key in clear text to a remote host unless plaintext LAN is allowed (#4131) — thanks @DeryFerd!
- Segmented model downloads pin public destinations with bounded DNS checks across redirects, ranges and proxies, preventing private-network fallback and CDN credential leaks (#4132) — thanks @DeryFerd!
- Cloud sign-in now rejects a discovery document whose token or revocation endpoint is not on the issuer's own origin (#4133) — thanks @DeryFerd!
- Transcriptions can save clean plain text and timed SRT captions, with an explanation when subtitle timings are unavailable (#4135)
- Stopped playback errors from incorrectly claiming that generated audio files are missing (#4138) — thanks @AAYUSH-CODER1904!
- Android model downloads resume through VPNs while preserving Wi-Fi policy and saved ranges (#4140) — thanks @JerryTimi!
- Make the installed-model licence action open its terms and acknowledgement dialog — thanks hugotriunfa18 on Discord! (#4147)
Contributors
- New code contributors since v0.5.9-beta.2: @DeryFerd and @vandyahmad24.
- Code since v0.5.9-beta.2: @debpalash, @DeryFerd, @vandyahmad24.
- Code since v0.5.8: @debpalash, @DeryFerd, @ItzSudhan, @velixio, @b84330808, @blutuxyz, @kevin9327, @RerankerGuo, @hkarmoush, @vandyahmad24.
- Reports, requests and performance feedback: @ftlcustomerservice73-ui, @mrkeshi, @cf3901646, @MaverickKB, @xabiherdz-svg, @nntloc, @AAYUSH-CODER1904 and @velixio.
- Transcriptions and Dubbing plain-transcript request: @shivamdhiman999111-source; Android VPN report: @JerryTimi; cleanup proposal: @themrbeasley; Windows crash report: @renaldyramadani123-code.
- VoxCPM2 installation report: vinhson on Discord, from the original complaint; unresolved, with no GitHub identity attributed.
- Installed-model licence-action report: hugotriunfa18 on Discord, from the original complaint; no GitHub identity is attributed.