Find the waiting stage#
If no draft has appeared after recording, check local model readiness and the capture/transcription state. If a raw draft exists but optional cleanup is waiting, inspect service configuration and account readiness. If the draft is correct but sending is pending, use the receipt guide.
Try a smaller ordinary-speech recording. Note the duration, input setup and whether the model was newly preparing. Avoid starting overlapping long attempts that make results harder to identify.
Keep claims realistic#
The reviewed materials do not establish a latency guarantee for every Mac, recording length or language. Universal packaging includes Intel, but not every Intel machine has a real performance test.
If a wait persists, retain the useful text and the visible state. A report should identify which stage is slow, rather than assuming every delay is a network problem. Local transcription does not depend on the optional refinement service, while later AI task work may depend on network and model capacity.