Changelog

What's new.

Every public release of ShoutFlow, newest first. Subscribe via RSS to follow along.

2026.08.011 AUGUST 2026LATESTstable

ShoutFlow 2026.08.0

Choose the spoken language for better recognition, keep dictated text in its original language, and paste reliably after processing.

new
  • Choose a spoken language in Settings, with automatic detection still available.
improved
  • Audio input now recovers more reliably after route changes.
fixed
  • German and other non-English dictation now stays in the dictated language; Auto-Paste returns text to the app where recording started.
2026.07.324 JULY 2026stable

ShoutFlow 2026.07.3

More reliable setup, permissions and local model downloads, with a corrected Try It experience.

new
  • Rebuilt first-run setup and permission repair for Microphone and Accessibility.
improved
  • Large v3 Turbo now uses the compressed ~1.05 GB model with coherent download progress, speed and ETA.
fixed
  • Resolved signing and keychain recovery, duplicate model loading, repeat multi-gigabyte downloads, and Try It text alignment.
2026.07.221 JULY 2026stable

ShoutFlow 2026.07.2

Automatic seven-day trials, clearer onboarding, flexible BYOK setup, and improved support diagnostics.

new
  • Start a seven-day free trial automatically on first launch.
  • Choose on-device processing or BYOK with OpenAI, OpenRouter, DeepSeek, Kimi, Gemini, xAI, or a custom provider.
  • Optionally open ShoutFlow automatically at login.
improved
  • See the exact trial expiry date, remaining time, and purchase path throughout onboarding.
  • Support requests now include clearer purchase and licence identifiers plus useful app and Mac details.
security
  • Strengthened device-level trial integrity checks.
2026.07.120 JULY 2026stable

ShoutFlow 2026.07.1

Automatic updates, secure production licensing and release channels, privacy-safe transcription diagnostics, and polished onboarding and settings.

new
  • Added automatic update checks with verified, rollback-safe installation.
improved
  • Added secure production licensing, trials, and separate Production and Developer build channels.
fixed
  • Improved onboarding, permissions reporting, recent transcription history, and floating recording controls.
security
  • Removed transcript content and raw provider responses from diagnostics.
0.9.417 JULY 2026beta

Control Island, redesigned

The bottom capsule now uses the ShoutFlow mark itself as the state indicator, shrinks when idle, and drops the separate status badge entirely.

new
  • Logo-as-state indicator: white idle, pulsing red recording, fast blue processing.
  • Solid red-dot start button replaces the old waveform control.
improved
  • Island width now hugs content per state: compact at rest, expanded while recording.
fixed
  • Collapse control no longer overlaps the timer on narrow displays.
0.9.210 JULY 2026beta

Live transcription preview in the overlay

The recording overlay now expands to show a live preview of your words as WhisperKit transcribes them, so you can see the take before it lands.

new
  • Streaming preview row below the overlay title, up to three lines.
improved
  • Preview scrolls smoothly and clamps instead of jumping.
fixed
  • Overlay no longer flickers when switching between apps mid-recording.
0.9.02 JULY 2026beta

Bring your own key: five providers at launch

Connect OpenAI, OpenRouter, Gemini, Kimi or xAI with your own API key. Keys live in the macOS Keychain and requests go directly from your Mac to the provider.

new
  • Provider settings with Keychain-backed key storage.
  • Per-app tone mapping, so Slack gets casual and Mail gets formal.
improved
  • Clearer fallback to local mode when a provider request fails.
security
  • Keys are never transmitted to ShoutFlow servers; verifiable with any network inspector.
0.8.020 JUNE 2026beta

Local mode: WhisperKit plus MLX, entirely on-device

The default pipeline now transcribes with WhisperKit and cleans up with MLX on Apple Silicon. Nothing content-bearing leaves the Mac.

new
  • On-device transcription via WhisperKit on the Neural Engine.
  • MLX clean-up pass: filler removal, punctuation, grammar and tone.
improved
  • Transcription starts while you are still speaking, with zero network latency.
5 JUNE 2026

First private beta

ShoutFlow exists outside a screen recording. Floating overlay, global hotkey, and text insertion into any Mac app.

new
  • Global hold-to-talk hotkey with floating recording overlay.
  • Clean text inserted at the cursor in any macOS app.
fixed
  • Everything, repeatedly. It was the first build.