Capture

ThoughtSift captures a thought as audio and transcribes it. Several transcription engines are available, and you can also type directly.

Capture entry points

On iOS, you can start a capture from the Action Button, a home-screen widget, or the app. The screen can lock while you are still talking, and recording carries on until you stop.

Transcription engines

• Apple Speech Analyzer — the default on-device engine where iOS 26 supports it.
• On-device Whisper (WhisperKit) — the default on-device engine otherwise. The model downloads to the phone; the English model is used for English recordings.
• Cloud transcription (Google Gemini) — optional. You choose it in Settings.

When transcription runs on-device, your voice audio does not leave the device. ThoughtSift sends the text transcript to build your sift.

If Whisper can’t run (the model isn’t downloaded yet, or the recording isn’t in English), ThoughtSift asks before sending anything: Send audio to cloud or Keep audio on this iPhone. If you keep it, the recording is saved on the phone so you can retry. With cloud transcription, recordings up to about 20 minutes and 14 MiB go straight from the app to Google’s Firebase and Gemini services. Longer or larger recordings (over about 20 minutes, or over 14 MiB even if short) are uploaded to your account’s private storage on ThoughtSift’s backend first, then transcribed there with Gemini. The limits are 30 minutes and a 32 MiB upload.

Meeting mode

Meeting mode asks the transcriber for speaker labels and structured fields, so multi-speaker captures come back organised.

Text input

If you’d rather type, the text-input path skips recording and feeds your text straight into content intelligence.

Manual review

You can enable a manual review step to edit the transcript before a diagram is generated, so the visual reflects exactly what you meant.