Troubleshooting common issues
Solutions to the issues that come up most often, gathered in one place.
"FFmpeg required" or "Setup required" will not go away
Open Settings → Dependencies, see which core tool shows Not installed and click Download & Install. The banner lists exactly what is missing. A tool cannot be installed while a job is using it — wait for the job or stop it. If a download fails, try again: FFmpeg is fetched from several sources in turn, and the message lists each one that failed.
Transcription produces no subtitles
Voice Activity Detection can filter out quiet speech it mistakes for silence. Turn it off in Settings → General and try again. Also check the right speech model is chosen on Transcribe, and the language matches the audio.
Transcription repeats the same phrase, or cues all start on the same beat
Whisper can loop on music or noise. Turn on Reduce hallucinations in Settings → General. If many cues share the same millisecond and last about two seconds, that is the effect of a vocabulary hint on timing; the same switch fixes it by ignoring the hint.
Transcription is slow, or the GPU is not used
Check which Whisper build is installed on the Whisper.cpp card in Dependencies — only a CUDA, Vulkan or Metal build can use the graphics card. A GPU build whose graphics support cannot start quietly runs on the CPU; reinstalling Whisper from Dependencies often fixes a CUDA build. Qwen3-ASR and Parakeet always run on the CPU. A file marked Waiting is queued for its share of the processor or card. Raising Videos transcribed at once helps CPU runs only.
Translate or AI Clean Up is disabled
Translate needs something to pay for it: a DeepSeek key saved in Settings → AI Providers together with an activated license, or a connected credits account with credits left. Check Pay with. AI Clean Up, Suggest vocabulary and Write context with AI need your own key and a license; credits cannot pay for them. Translation that stops with an out-of-credits or rejected-key message keeps what it finished and leaves the rest blank.
License will not activate
Check the key for typos (format WCSL-XXXX-XXXX-XXXX-XXXX-XXXX). If it is right you may be at your device limit — deactivate an old device from the website’s license portal and try again. See Managing your devices.
Extract gave me a .sup or .sub file, not an .srt
That track is made of pictures (Blu-ray PGS or DVD VobSub). Extract saves it as it is. Turn it into text on Convert, which reads it with built-in OCR.
OCR is slow or the first conversion needs the internet
The first OCR run downloads its recognition model, so it needs a connection once. After that OCR runs offline: on the graphics card through DirectML on Windows when available, otherwise on the CPU — Macs always use the CPU. Wrong characters on low-resolution scans can be improved with Clean Up’s Fix OCR errors option.
YouTube Transcript does nothing or reports a rate limit
YouTube throttles devices that make many requests quickly. Wait a few minutes and try again. If it persists, update yt-dlp on the Dependencies screen — YouTube changes often.
No waveform, or playback controls are greyed out
You opened a subtitle-only file with no video. Use Open video in the banner to load the matching one.
Where are my files saved?
By default next to the source video or subtitle file. Extraction, transcription and Save try that folder first; if it cannot be written to, a save dialog lets you choose another place.