The transcript is empty, silent on one side, or in the wrong language

Only my side is transcribed

The system-audio channel was not captured. On macOS this is the Screen & System Audio Recording permission, which unsigned builds can lose after an update. Toggle it off and on in System Settings → Privacy & Security, then record a short test. On Windows and Linux (early access) it can also be a capture problem; send diagnostics.

Only their side is transcribed

The microphone permission, or the wrong input device. Check your operating system’s microphone permission (on macOS, System Settings → Privacy & Security → Microphone) and Settings → Capture → Microphone in Recordist (a USB headset that was unplugged mid-call is the classic cause).

Nothing at all

Open the meeting and look at the waveform. Flat lines on both channels mean nothing was captured (permissions). Waveforms with no text mean transcription failed: check that the model finished downloading under Settings → Transcription, and use Re-transcribe on the meeting.

Wrong language

Automatic language detection decides from the first seconds of speech. A greeting in one language before a meeting in another can send it the wrong way, and English-only .en models cannot transcribe anything else. Pin the language under Settings → Transcription → Language and click Re-transcribe. Mixed-language meetings work best with the default large-v3-turbo and auto-detect.

Words are there but garbled

Try the default model if you are on tiny or base, turn on Denoise under Capture, and move closer to the microphone. Voice-activity detection is on by default and trims silence, not speech; if it clips the start of sentences, turn it off for that meeting and re-transcribe.