Troubleshooting

Locate the failing stage and change one setting at a time. Do not delete the entire user-data folder to troubleshoot; it can contain meetings, terminology, voice material and credential-protection files.

Installation or startup

• Confirm the package came from the official download route, then check its version, source and signature information.
• Review disk space, permissions and security-software messages. Do not disable protection or exclude the whole directory.
• Follow the app's environment checks for missing components. Check connectivity before retrying a component or model download.
• Exit normally and reopen. If the app stays blank or cannot reach its local service, retain the error details and contact support.

Do not indiscriminately stop other applications' Python processes or delete port, token or database files based on old paths.

No recognized text

1. Check whether you selected microphone, system audio or application audio.
2. Confirm that source has sound, is unmuted and has permission.
3. Check the active ASR binding and model readiness.
4. For APIs, check credentials, balance and connectivity. For local engines, check models and hardware.
5. Retry with a short sample and note any app error.

If the wrong language is recognized, check source-language selection. Quiet, noisy or intermittent audio also affects results.

Original text works, translation does not

• Check the active translation binding, target language and service status.
• If you edited a prompt, compare its variables with those listed by the template. Restore the app's default template if needed and retry.
• Check whether recognition already changed the input and whether glossary terms conflict.
• Inspect provider quotas, request limits and network errors.
• Review important numbers, names and commitments against the original.

See engine setup for prompts and configuration. This guide does not rank external models by fixed speed or quality.

TTS produces no audible output

Distinguish a synthesis failure from audio routed elsewhere.

1. Check the TTS engine, voice and target language.
2. Check local playback versus virtual-microphone output.
3. Review the output device, system volume, mute state and app errors.
4. For meeting output, check the meeting app's input endpoint too.

Follow the virtual microphone guide. For echoes, inspect whether output is being captured as input again.

No knowledge-base results

Check enabled entries, category selection, embedding configuration and completed vectors. Changing the embedding model can require regeneration. Try a representative question in the search-test dialog.

Results depend on thresholds, wording and entry quality. The current workflow uses categorized Q&A and compatible JSON imports. See knowledge retrieval.

Activation, devices and renewals

• Invalid key: check copied text, order status and possible revocation.
• Device limit: release an unused device in device management.
• Expired access: review status and available actions in subscriptions.
• Network-check failure: check HTTPS connectivity, system time and your organization's firewall policy. The license service domain is license.crossmeet.com.

Do not post a complete key in public feedback. See activation and devices.

Performance and models

Check the actual model, hardware and concurrent tasks. Reduce competing work and select a configuration your device supports, then retest the same sample. Initial model loading can be slower than later calls.

Do not apply one GPU driver, memory or segmentation requirement to every model. Use the installer and app's checks for the selected engine.

History and migration

The account website does not automatically sync local history. Use available export or backup features to move content. Before overwriting, cleaning models or uninstalling, review the scope and back up.

Local data normally lives under the current Windows user's %APPDATA%\CrossMeet\ directory. Do not assume every database is directly inside that folder.

Still need help?

Email support@crossmeet.com with:

• CrossMeet and Windows versions.
• The affected workflow, engine and model.
• The exact error, time and reproduction steps.
• Checks you have already tried.

Send logs and screenshots only after reviewing and removing sensitive content. Do not send passwords, API keys, license keys, original customer recordings or an unchecked complete diagnostic bundle.