Veyro 1.10.37

[1.10.37]

Fixed

• Capture / Status UI: Mic monitor IPC no longer holds the controller mutex while waiting on the audio pipeline, which could stall PTT release and freeze the UI during capture.
• PTT release: Deferred finish_ptt_release when the controller is busy; capture format is queried before taking the audio lock on pipeline start.
• Dictation state: Injection from Listening transitions through Processing instead of an invalid direct jump to Injecting.

Veyro 1.10.35

[1.10.35]

Fixed

• Windows release UI: devUrl moved to tauri.dev.conf.json only (not embedded in release builds), Vite uses relative asset paths, and the release script verifies embedded UI before publishing.

Veyro 1.9.32

[1.9.32]

Added

• Tools — Audio в†’ SRT: Optional speaker diarization (local local-diarization / polyvoice), rename/merge speakers, and WebVTT export with <v Name> voice tags alongside SRT.
• Tools — Vocal separator: Multi-stem instrument split (HT-Demucs 6-stem) with the same model-download confirm + progress UI as other downloads.
• Tools — Dictation transcripts: Dedicated tool window for reviewing / exporting dictation transcript history.
• Build: local-diarization feature wired through resolve-local-features.ps1 (opt out with VEYRO_DISABLE_DIARIZATION=1).

Changed

• Windows: Tool windows (catalog and individual tools) stay always-on-top like the main Veyro window.
• Tools — Audio в†’ SRT: Diarization toggle disables immediately when the feature is missing from the build (no on/off flicker).
• Tools — Audio в†’ SRT: Preview seek no longer flashes speaker badges or snaps the scrubber while dragging.

Fixed

• Tools — Vocal separator: Instrument multi-stem pass shows the usual processing progress bar while running.

Veyro 1.9.27

[1.9.27]

Added

• Injection: Soft line breaks (Shift+Enter) so paragraph newlines in chats do not send the message; toggle in settings (on by default). Hard Enter remains available for Word-like editors and for the Enter trigger phrase.
• Tools — Voice files: Save transcript as DOCX; auto-watch controls moved to a compact sidebar footer with expert options in a popup.
• About: Collapsible Models list of local downloadable models; third-party list curated to key technologies only.
• README: Promo screenshot banner before Quick start.

Changed

• Capture: Microphone / Microphone + app output are icon toggles beside the mic device picker; the app picker hides in mic-only mode.

Fixed

• AI rewrite: Custom skill mode no longer fails with В«custom skill is missingВ» when the selected .md has only frontmatter / empty body — falls back to Optimization, excludes empty skills from the picker, and rejects empty catalog/import installs.
• Skill catalog: Locked/unpaid catalog skills (access: locked) return empty content from the API while the website still shows a preview — install is now rejected with a clear error instead of writing a frontmatter-only file that breaks AI rewrite.
• Skill catalog: Partner-install flow per docs/veyro-catalog-partner-install.md — parse catalog_id + install_token from veyro:// links and download via /api/catalog/{id}/partner-install (fixes purchased skills that looked “locked” without a token).

Veyro 1.9.25

[1.9.25]

Added

• Tools — Voice files: Auto-watch folders for messenger voice downloads; shared processing queue with history index, tray controls, and OS notifications.
• Tools — Voice files: Wider tool window with a collapsible mini file browser (real paths, status, remove-from-index without deleting the file).
• Capture: Optional Microphone + app output mode (Windows WASAPI process loopback) — mixes selected app audio with the microphone for dictation.

Changed

• Tools: Subtitles and vocal/instrumental tools keep the previous vertical layout; only voice-file processing uses the sidebar browser.
• Windows: Tool windows use normal z-order (not always-on-top); the main Veyro window stays on top.

Fixed

• Tools — Voice files: Selecting a file in the sidebar shows its processed text in the result field.

Veyro 1.9.21

[1.9.21] — 2026-09-28

Added

• Tools: Split vocals / instrumental — offline stem separation with downloadable ONNX models, progress UI, and preview/download for vocals and instrumental stems.
• Separation models: Google ReFormer (quality) and Google ReFormer (speed) (musetric Mel-Band RoFormer, host STFT + CPU ONNX); Hybrid Transformer Demucs (StemSplitio htdemucs-ft vocals).

Fixed

• Separation: STFT time dimension is read from the ONNX graph so frame count matches the loaded model; iSTFT uses correct realFFT normalization.

Veyro 1.9.4

[1.9.4]

Added

• Tools: Full localized error text in voice file and Audio SRT result areas; music/silence-only transcripts show an info message instead of raw [Music] tags.
• Tools: Audio SRT advanced subtitle settings use sliders with sensible ranges and live value labels.

Changed

• Capture: Recognition language list sorted by visible name (locale-aware); Auto stays first.
• Tools: Decode/transcription progress uses staged backend percentages and smoother UI updates.

Fixed

• Capture: After PTT injection, leaving the target field clears the pinned injection target so the next dictation goes to the focused field.
• Tools: Progress bar fill visible again; decode stage no longer jumps to 100% on WAV/preprocess.

Veyro 1.9.3

[1.9.3]

Added

• Tools: Audio в†’ Subtitles (SRT) — transcribe files with segment timestamps, advanced subtitle layout (smart split, line limits, reading tail), preview timeline, export .srt.
• Tools: When recognition language is Auto and STT returns empty, prompt to pick a language and retry (Audio SRT and voice files).
• STT: Auto language detection for local Whisper; fixed-language and post-process fallback policy; nullable settings patch so Auto persists from the UI.

Changed

• Audio SRT: Advanced settings use per-field ? hints; hide pause-based split when Smart split is on.

Fixed

• Settings: Recognition language select (including switching back to Auto) saves reliably.
• Tools: Vertical scroll in tool windows when content overflows.

Veyro 1.8.94

[1.8.94]

Fixed

• PTT hold mode (Windows): Stop spurious release/press loops while the hotkey chord is held — hook edges only, no GetAsyncKeyState reconcile on every key event.

Veyro 1.8.91

[1.8.91]

Added

• Tools: Embedded tools panel and voice file transcription window — queue audio files, same local/cloud STT and text pipeline as dictation (including AI optimization and custom skills); copy results to clipboard.
• Voice files: Decode common formats (WAV, MP3, OGG/Opus, FLAC, M4A, …); progress percent during decode and local Whisper inference; EN/RU stage labels.
• UI: Tools entry in the status bar (wrench icon); link from Standard (homemaker) settings.

Fixed

• PTT / capture: Do not block dictation while a voice file job holds the tools lock only on the tool side; restore audio callbacks after PTT errors; hotkey capture button uses mousedown so assignment works in WebView2.
• Voice tool: OGG Opus via symphonia adapter; respect optimization / custom skill modes; local LLM load before rewrite; Silero TE tokenizer tensor shapes; strip common Whisper hallucination phrases.
• README: Document tools, Sherpa/Whisper STT families, and voice file workflow.