This release adds 9 notable features for engineering teams evaluating rollout.
✓ No known CVEs patched in this version
Summary
AI summaryPerformance improvements keep dictation streaming, writing quality adds proper punctuation, and Live transcript shows up to two readable HUD lines.
Changes in this release
| Type | Severity | Summary | CVE |
|---|---|---|---|
| Feature | Low |
Adds sentence‑ending punctuation to complete thoughts in dictation output. Adds sentence‑ending punctuation to complete thoughts in dictation output. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Feature | Low |
Preserves final full stop in longer Terminal prose while keeping short commands model‑free. Preserves final full stop in longer Terminal prose while keeping short commands model‑free. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Feature | Low |
HUD now shows up to two readable lines selected on word and sentence boundaries. HUD now shows up to two readable lines selected on word and sentence boundaries. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Feature | Low |
Whisper previews use the warmed model that produces the final transcript, preventing state corruption. Whisper previews use the warmed model that produces the final transcript, preventing state corruption. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Performance | Medium |
Reuses Qwen prompt prefix during dictation, avoiding repeated cost. Reuses Qwen prompt prefix during dictation, avoiding repeated cost. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Performance | Medium |
Cancels stale preview and cleanup work, restarts sidecar if MLX inference wedges. Cancels stale preview and cleanup work, restarts sidecar if MLX inference wedges. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Performance | Medium |
Stops repeated full‑recording audio copies during Whisper previews. Stops repeated full‑recording audio copies during Whisper previews. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Performance | Medium |
Keeps long Terminal dictations streaming even when the first segment is short. Keeps long Terminal dictations streaming even when the first segment is short. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Performance | Medium |
Pauses hidden HUD animations, resulting in 0.0% idle CPU across 12 samples. Pauses hidden HUD animations, resulting in 0.0% idle CPU across 12 samples. Source: llm_adapter@2026-07-17 Confidence: high |
— |
| Bugfix | Medium |
Fixes clear grammar and agreement issues in Qwen output without altering meaning. Fixes clear grammar and agreement issues in Qwen output without altering meaning. Source: llm_adapter@2026-07-17 Confidence: high |
— |
Full changelog
Velora 0.4.5 makes dictation faster without swapping or shrinking either configured model.
Performance
- Reuses the exact Qwen prompt prefix prepared while you speak instead of paying that cost after release.
- Cancels stale preview and cleanup work, and restarts the sidecar if MLX inference hard-wedges.
- Stops repeated full-recording audio copies during Whisper previews.
- Keeps long Terminal dictations streaming even when the first segment is short.
- Pauses hidden HUD animations; the installed app and loaded sidecar measured 0.0% idle CPU across 12 samples.
Writing quality
- Complete thoughts now receive sentence-ending punctuation.
- Longer Terminal prose keeps its final full stop while short commands remain model-free and command-safe.
- Qwen fixes clear grammar and agreement issues conservatively without paraphrasing or changing meaning.
- Names, numbers, technical terms, and dictated details stay protected by the divergence checks.
Live transcript
- The HUD shows up to two readable lines selected on word and sentence boundaries.
- Whisper previews use the same warmed model that produces the final transcript and cannot corrupt final state.
The production models remain mlx-community/whisper-large-v3-turbo and mlx-community/Qwen3.5-4B-MLX-8bit.
Verified with 233 Python tests, 69 Swift self-checks, six exact-model quality fixtures, an installed end-to-end audio run (1.298s stop-to-final), Developer ID signing, Apple notarization, stapling, and Gatekeeper assessment.
SHA-256: a5541af5be8570573fc9f416ec167a4a7737da650f578ec25780d78e627c1671
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About Velora
All releases →Related context
Related tools
Beta — feedback welcome: [email protected]