Troubleshooting common issues

Last updated • 7 min read

Ten of the most common evoglyph problems and how to fix them. Each entry has a symptom, the likely cause, and the fix steps. If your issue is not here, email [email protected] with the contents of ~/Library/Logs/evoglyph/ attached.

Nothing happens when I press the hotkey

Symptom. You press your dictation hotkey and evoglyph does not start listening. The menu bar icon does not change state.

Cause. evoglyph cannot observe global hotkeys without Input Monitoring permission. The most common version of this is that you granted Input Monitoring once, then a evoglyph update reset it (a known macOS quirk).

Fix.

  1. Open System Settings → Privacy & Security → Input Monitoring.
  2. Confirm evoglyph is in the list and the toggle is on. If it is on, toggle it off and back on.
  3. Quit and relaunch evoglyph (right-click the menu bar icon → Quit).
  4. Test the hotkey again.

If the hotkey is conflicting with another app's global shortcut, change it from evoglyph → Settings → Hotkey. The choosing-a-hotkey docs article has a list of common conflicts.

Text is not appearing in the app I am focused on

Symptom. evoglyph transcribes your speech (you can see it land in the dashboard or briefly in the menu) but no text shows up in the app you are focused on.

Cause. evoglyph injects text via macOS Accessibility APIs, which require Accessibility permission. Either it was never granted, or a macOS update reset it.

Fix.

  1. Open System Settings → Privacy & Security → Accessibility.
  2. Confirm evoglyph is enabled. If it is, toggle off and back on.
  3. Quit and relaunch evoglyph.
  4. Try dictating into a known-good target like TextEdit to confirm injection works at all. If TextEdit works but a different app does not, the target app may be sandboxed in a way that blocks Accessibility text injection — see the next entry.

Text lands in the wrong app

Symptom. You meant to dictate into App A, but the text appears in App B (often a window that briefly took focus during dictation).

Cause. evoglyph decides which app to inject into based on which app holds keyboard focus at the moment of injection. If a notification, a system dialog, or a brief background app steals focus during your dictation, the text follows focus.

Fix.

  1. Click directly inside the target app's text field before pressing the hotkey.
  2. Avoid switching windows while you are speaking. evoglyph injects when you release the hotkey.
  3. If a specific app keeps stealing focus (Slack notifications are a frequent offender), try quieting its notifications during dictation, or close it before dictating into another app.
  4. If the wrong-app pattern is repeatable in a specific app, email [email protected] with the app name and version — we keep a list of known-bad apps.

evoglyph will not launch (“App is damaged”)

Symptom. Double-clicking evoglyph produces a Gatekeeper dialog that says “evoglyph is damaged and can't be opened.”

Cause. Almost always an incomplete or corrupted download. evoglyph ships notarized and code-signed; a fully downloaded build does not trigger this dialog.

Fix.

  1. Move the existing evoglyph.app to the Trash.
  2. Re-download the latest build from evoglyph.com.
  3. Drag the fresh copy into Applications, then launch.
  4. If the dialog appears again, your network may be modifying the download (corporate proxies, content filters). Try a different network and re-download.

Symptom. evoglyph was running, then the menu bar icon disappeared. The hotkey no longer works.

Cause. evoglyph crashed in the background. macOS does not always show a crash dialog for menu-bar apps.

Fix.

  1. Relaunch evoglyph from Applications. The icon should reappear.
  2. Open Console.app, search for evoglyph in the Crash Reports section, and grab the most recent report.
  3. Email it to [email protected], or — if you would rather it go automatically next time — turn on crash reporting at Settings → Privacy. Crash reporting is off by default; the next crash will be sent to us via Sentry with audio/transcripts/PII scrubbed.

evoglyph's local debug log lives at ~/Library/Logs/evoglyph/debug.log and is also useful to attach.

First-run model download is stuck

Symptom. First launch shows a model-download progress indicator that is at 0%, frozen, or hanging well past a few minutes.

Cause. One of three things: a firewall or content filter is blocking the model host, you are out of disk space, or the connection died and evoglyph is waiting on a retry.

Fix.

  1. Click Cancel in the download dialog, then Retry from the menu bar icon → Models.
  2. Confirm you have at least 3 GB of free disk space (the transcription and cleanup models combined are around 2 GB).
  3. If you are on a corporate network, try a personal hotspot or home network — content filters that block AI model hosts are increasingly common.
  4. If the retry fails repeatedly, email [email protected] and attach ~/Library/Logs/evoglyph/debug.log.

Microphone permission keeps reverting

Symptom. You grant microphone access to evoglyph, dictation works for a while, then suddenly evoglyph cannot hear you and macOS says permission is denied — even though the toggle in System Settings still looks on.

Cause. A known macOS bug where the TCC database (the system that tracks permission grants) gets out of sync between its on-disk state and what apps actually see.

Fix.

  1. Reboot your Mac. This is the cleanest fix and resolves it immediately.
  2. If you cannot reboot right now, open System Settings → Privacy & Security → Microphone, toggle evoglyph off, then back on, then quit and relaunch evoglyph.

Transcription is wrong on technical terms or names

Symptom. General prose comes out fine but technical jargon, product names, or specific people's names get mangled.

Cause. The transcription model has not seen these words often enough in its training data. This is normal for any speech-to-text system.

Fix. Add the words to your custom vocabulary from evoglyph → Settings → Vocabulary. Vocabulary terms are passed to the local transcription engine as biasing hints. See the vocabulary docs article for tips on writing entries that actually help (use phonetic spellings for tricky names, group related terms, etc.).

evoglyph is using too much RAM

Symptom. Activity Monitor shows evoglyph holding around 2 GB of resident memory.

Cause. The MLX cleanup model holds about 2 GB resident while loaded so it can run instantly when you finish a dictation. This is normal — loading and unloading the model on every dictation would add seconds of latency.

Fix. If you would rather trade cleanup for the RAM, disable AI cleanup at evoglyph → Settings → Cleanup. Transcription still works; you just get raw transcripts. RAM usage drops to a few hundred megabytes.

Auto-update is not working

Symptom. A new evoglyph release is out (you saw the changelog) but evoglyph is not offering to update.

Cause. evoglyph uses Sparkle to check evoglyph.com for updates on a schedule. If Sparkle's request is blocked, or if you disabled the check, you will not see new versions.

Fix.

  1. Confirm auto-update is enabled at evoglyph → Settings → Updates.
  2. Click Check for updates now to force a check.
  3. If the check fails, your network may be blocking evoglyph.com — try a different network.
  4. As a last resort, download the latest build manually from evoglyph.com and drag it over the existing app in Applications. Your settings, vocabulary, and history are preserved.

Still stuck?

Email [email protected] with:

evoglyph's local debug log does not contain audio or transcripts; it logs lifecycle events (launches, hotkey presses without keystroke content, errors) and is safe to share. [verify: confirm log redaction policy with the diagnostics work in evoglyph#614]

Was this helpful?
Related articles