A running log of new features, quality improvements, and release notes as Mallo gets better for macOS voice workflows.
The dictation HUD appears more reliably over fullscreen apps, reacts more smoothly when you start speaking, and behaves better when you toggle it quickly.
Mallo 1.2.19 adds a repair and reinstall flow for Qwen, so you can refresh the setup directly from Models when it needs a clean start.
Mallo 1.2.17 makes the first Qwen dictation after an update behave more cleanly by showing a short preparing state instead of failing too early.
Mallo 1.2.16 fixes another Qwen upgrade case, so older managed installs keep working after you update instead of looking missing.
Mallo 1.2.15 restores existing Qwen setups after upgrade, so installed runtimes keep working without asking you to reinstall.
Mallo 1.2.14 makes Qwen setup more reliable, fixes the Models-page install path, and gives slow downloads more room to finish cleanly.
Mallo 1.2.1 improves HUD control, stabilizes repeat Qwen dictation attempts, and adds cleaner synthetic dictation testing paths.
Mallo now lets you pick the audio input inside the app, and dictation holds up better even when your Mac is already playing sound.
Mallo now does a better job of surfacing permission issues, guiding recovery, and smoothing the fallback paths around clipboard and notifications.
We improved Qwen capture readiness and tightened hotkey behavior so dictation starts more reliably in the moments you actually reach for it.
Mallo now handles no-focus dictation more gracefully, so your words are less likely to disappear when there is nowhere obvious to type.
Mallo now gets Qwen ready much faster on cold start, improves the default model path in onboarding, and behaves better over fullscreen apps.
Use a single modifier key like Command, Control, Shift, or Option for hold-to-talk or toggle dictation in Mallo.
Mallo 1.2.0 brings stronger multilingual dictation, a cleaner models experience, and managed Qwen setup on macOS.
Install Qwen3-ASR-1.7B-8bit from the Models page with a managed runtime path instead of a manual Python setup.
Qwen now lives in the main model list with cleaner setup guidance, better ordering, and a more native Mallo experience.
We simplified the tray menu and broadened speech model coverage so everyday dictation feels more direct.
We simplified diagnostics helpers and core macOS app plumbing to make the app more stable and easier to evolve.
Mallo now uses a more consistent model selection flow across navigation, tray entry points, and model management.
Mallo now includes Parakeet as a speech provider, adding stronger multilingual model options alongside Whisper.
Shipping continuously
We’ll keep logging multilingual dictation, model setup, licensing, and workflow improvements here.
