Troubleshoot slow transcription
Diagnose long local processing, stalled file jobs, unavailable GPU acceleration and slow Cloud requests.
Last updated
Identify the processing path
Local Mode speed depends on the selected model, recording length, CPU, GPU, drivers and other open applications. Cloud depends on the internet connection and service availability.
First actions
- For file jobs note whether the row says Preparing or Transcribing. For dictation also check whether translation or a dictation Mode is enabled; these add processing after recognition.
- For Local Mode install the latest graphics driver and restart Windows.
- Try a lighter installed model such as Small q8 or Turbo q5.
- If CPU is selected, test a reasonable thread count while keeping Windows responsive.
- Close other demanding CPU, GPU and disk workloads then test a short clear recording.
- For Cloud confirm the internet connection and retry a short recording.
Long-file recovery
WhisperARC splits long media into smaller sections. If one section times out, it is tried once more. Completed sections can be reused after an interruption. Retry the failed row instead of adding the source again.
Collect useful details
- WhisperARC and Windows versions
- Selected Local, Cloud or OpenAI KEY source
- Selected transcription model and CPU or GPU choice
- CPU model, RAM amount and every graphics adapter
- Graphics driver version
- Source duration, job stage and approximate elapsed time