Mallo 1.2.29
Mallo 1.2.29 makes Whisper setup reliable again and fixes dictation that could remain pending after recording.
Mallo 1.2.29 makes Whisper setup reliable again. You can install a Whisper model during onboarding and start dictating without a separate runtime download, and dictation no longer stays pending after recording.
Release Highlights
- Whisper is ready with one model download. Everything else Whisper needs now travels with Mallo.
- Incomplete downloads never appear ready. Mallo verifies the finished model and retries temporary download failures before activating it.
- Dictation finishes instead of staying pending. Recording and transcription now stop cleanly when you finish or cancel.
- Other local models stay untouched. Clearing Whisper downloads does not remove your Qwen or Parakeet setup.
What Changed In Real Use
Previously, changes outside Mallo could make Whisper setup fail even after it had worked for months. In some cases, a manual workaround got the model installed but left dictation waiting indefinitely after recording.
This release keeps the Whisper components Mallo needs inside the app and verifies each model before it becomes available. Temporary network errors are retried, interrupted files are not treated as complete, and a stalled transcription cannot leave dictation running forever.
Why This Patch Matters
Choosing Whisper should be enough to use Whisper. Whether you are setting up Mallo for the first time or returning to a model you already use, the app now checks that the model is complete and keeps the rest of your local speech setup separate.
Related updates
Keep reading
Model selection is now unified across the app
Mallo now uses a more consistent model selection flow across navigation, tray entry points, and model management.
Diagnostics and app plumbing got simpler
We simplified diagnostics helpers and core macOS app plumbing to make the app more stable and easier to evolve.
Managed Qwen setup inside Mallo
Install Qwen3-ASR-1.7B-8bit from the Models page with a managed runtime path instead of a manual Python setup.
Shipping continuously
Follow each release as Mallo gets sharper.
We’ll keep logging multilingual dictation, model setup, licensing, and workflow improvements here.