If something goes wrong in VokeLine, find the section closest to the message or symptom and work through it in order. Save or copy any text still visible before troubleshooting. Reinstalling, clearing data, and repeatedly regenerating should not be your first steps.
Installation and updates
The installer will not open
- Check that the package came from the download page, matches your platform and architecture, and has finished downloading.
- Download the currently available version again. If the system blocks it, verify its source and follow the normal system installation and opening prompts.
- On macOS, check that the app is in Applications. For a Linux AppImage, allow it to run as a program.
- If your platform or architecture is not offered, contact support rather than using another platform's package.
After installation works, prepare a model. The download page shows which platform and architecture packages are available now.
Update checks fail or show no new version
Open Settings → Software updates, check the connection, and try again. If no update is available, continue with your current version. Keep the working app if downloading or installing an update fails, then follow the retry prompts. Finish active recordings or file tasks before updating.
Models and downloads
A downloaded model is still missing
Open Models, confirm task support and completed file checks, then select Use. Live input needs live-transcription support; file tasks need file-transcription support. Check that the appropriate default indicator appears. See Model management.
Downloads pause, fail, or do not pass verification
- Reconnect and resume a paused download. Check disk space against the listed download size.
- Use the retry control after a download failure. Redownload incomplete, damaged, or unreadable files as prompted.
- To delete and redownload a default model, switch to another usable model first and finish active transcription before removing that version.
For insufficient memory, close memory-heavy apps or choose a model your device can run. If the processor does not meet the model's requirements, select a suitable model. Repeated downloads will not fix unsupported hardware.
Downloads do not work offline
New downloads and resuming downloads need internet. Installed, usable local models can still transcribe recordings or files. AI, search, and cloud operations need a connection and should be checked separately.
Microphone and recording
No device is found or no sound is detected
- Connect a microphone and check physical mute and system input volume.
- Reopen the input-device list in Settings → Voice and shortcuts and select an available device. Do not keep using a disconnected selection.
- Run Test microphone and speak near the device. It finishes within 10 seconds without saving the test recording.
- If permissions are missing, open system microphone settings. Test again after allowing access; reopen the app if the device is still unavailable.
Try a short recording once sound is detected. See Check your microphone for entry points and platform settings.
Text is missing or inaccurate
Check that the model supports the spoken language and that the selected microphone works, then reduce background noise and try again. Live recording detects the audio language automatically; file transcription lets you select it when the model supports language selection. Changing the interface language does not change speech recognition. To explain, summarize, or extract key points in another language, choose the corresponding AI action and set its output language. The audio-language setting tells the model what language to recognize.
Shortcuts and Direct mode
A shortcut does nothing
- Open Settings → Voice and shortcuts and check whether registration succeeded or a conflict/unavailable warning is shown.
- Choose another combination if occupied, wait for saving, and test from your destination app.
- Linux may restrict global shortcuts; other apps or the system may take media buttons. Try the Home or floating microphone button to establish whether recording itself works.
- If the wrong mode opens, check whether you used that mode's specific shortcut or a button/media control that uses the default mode.
See Shortcuts and headset controls for platform defaults and setup.
Direct input fails or asks you to select a text field
Click an editable destination field and position the cursor before using the Direct shortcut. If VokeLine is waiting for a destination, select it as prompted. Keep that input location while recording; do not switch windows or move to another field.
Use Manage Direct mode permissions when required, and check Accessibility access on macOS. If the destination rejects insertion, use Editor mode to review and copy the text first, then troubleshoot Direct mode without losing the content.
Esc does not immediately hide the window
Esc first deals with the current popup or the switch between AI output and the original, then with collapsing. Use the specific hide or quit control when needed. Complete recording or save prompts before closing. See Stopping, collapsing, hiding, and quitting.
File transcription
Files cannot be added or transcription cannot start
Check for WAV, MP3, FLAC, OGG, M4A, or MP4; non-empty files of at most 2 GB each; and no more than 100 files per selection. Remove invalid entries and confirm a local file-transcription model and audio language are selected.
Select files again after renaming, moving, external changes, or an expired selection. If the extension is supported but decoding fails, check playback, the presence of an audio track, and whether the file is damaged.
A task fails, is interrupted, or has no text
- Read the task error and resolve model, file, or system-resource issues first.
- Retranscribe a failed task. Restart the needed task if the previous processing was interrupted.
- If a completed task has no recognized text, check for clear speech in the audio and try again with the correct audio language and an appropriate model.
Cancel task is available only in the applicable states. Clear completed tidies the task list, not history. See Transcribing audio files. Speaker separation is not supported in this version.
AI and credits
Sign-in or more credits are required
Check the active account and available credits in Account. If needed, open Account center, purchase, check the credit history, and return to the desktop app for updated account information. Run the action when credits and the AI service are available.
AI actions require sign-in, internet, and credits even after local transcription has succeeded.
The service is unavailable, input is too long, or processing fails
- Reload the service when it is unavailable, or try again later.
- Shorten or split oversized text, then check the operation's credit information again.
- For network failure, rate limits, or a busy service, retry later. Wait when the current request is still processing or being reconciled.
- Copy any visible result first. “Retry” attempts to recover the same request; “Run again” starts a new operation and may spend credits again.
See AI text actions for steps and language rules. If the credit change is unclear, check credit history and send feedback rather than repeatedly regenerating as a recovery test.
Sign-in and records
Google sign-in does not return to the app
Check that your default browser opens the login page and complete account selection within the same flow. Open VokeLine when the browser prompts you. After cancellation or timeout, start again from Account. Resolve network or Google-availability messages first, or use another method actually offered.
The session was replaced or cannot be restored
Another client's new sign-in to the same account replaces this desktop session. Sign in on the client you want to use and avoid repeated sign-ins on both sides. Use Retry after restoration fails, or sign in again if it is invalid. See Sign-in and sessions.
Local history seems missing after sign-in
Check whether the entry belongs to the guest, current account, or another account, and whether it was created on this device. Local history is separated by identity and device. Switching identities changes the visible partition without automatically migrating entries. For other devices, save to cloud in advance.
Cloud records are missing or will not save
Check that the account matches, the entry was explicitly saved, and you are online. Refresh and check Unclassified and the appropriate tag. Retain loaded content and retry after a later-page loading failure. Remove unneeded records or tags when limits are reached.
Retain the draft and compare the latest record after an editing conflict. Local and cloud deletion are independent. See Delete or stop saving records for the irreversible scope of clearing.
Get help or send feedback
Where to start: Open “Account center” from the desktop “Account” menu, then choose “Help & Feedback” on the website. Sign in if needed.
- Include the app version, operating system, feature, what you did, what you expected, and what happened.
- Include the exact message. Add the model name for model issues, format and size for file issues, or the relevant order/credit-entry time for credit issues.
- Hide unnecessary email addresses, personal text, and payment details in screenshots. Never send passwords, sign-in tokens, or service keys.
- Follow replies and add information to the same feedback thread rather than opening repeated reports for one issue.
Copy any visible text to a safe place before submitting. Keep current data if saving failed or processing is unconfirmed, resolve the specific message, and then retry.