KotoRoku records meeting audio, transcribes it on your device, and creates minutes. You operate it through the four tabs at the bottom of the screen.
Tap to begin recording. Recording never starts automatically upon detecting sound. Notifying participants and obtaining consent is your responsibility.
Ends the recording and moves on to organizing the transcript. It takes a moment before the basic information dialog appears.
Shows "Preparing", "Waiting for audio", "Recording", "Stopping", or "Paused". While audio is being recorded, "Recording" is always shown.
Shows the level your microphone is picking up. When the level exceeds the threshold, the silence timer is reset. Adjust the threshold in the Sound tab.
Recording stops automatically after a period of silence. Long pauses in a meeting can stop it unintentionally.
Shows the live transcript while recording, and the minutes once they are generated. Transcript text changes until it is finalized.
Enter the title, location, and participants, then tap "Generate" to create the minutes. If you cancel, you are asked whether to discard the recording. A discarded recording cannot be restored.
Recreates the minutes with different basic information. If you use a cloud AI, each generation incurs AI usage fees.
Shares the minutes document and, depending on your setting, the audio. How the recipient handles them is your responsibility.
Lets you recover a recording that ended abnormally last time. It cannot be used while recording or processing. Entries whose contents cannot be interpreted are shown in a state where they can be neither recovered nor discarded, as a safety measure.
Choose "On-device", "Local AI", or "Cloud AI". If you choose a cloud AI, the full transcript is sent to the AI provider you selected.
A concise version produced by Apple's on-device AI. Nothing is sent externally. Location and participants are not inferred and are recorded as "unknown".
Processing with Qwen3 or Gemma 3 on your device. Nothing is sent externally. Models are 1–3 GB, so downloading over Wi-Fi is recommended. This option is hidden on devices with limited memory.
Choose Gemini, ChatGPT, Claude, DeepSeek, or Custom. Retention periods and training-use conditions differ by provider. Be sure to read the notes on the selection screen. The recorded audio file itself is not sent.
Stored in the device keychain. API keys are never sent over unencrypted HTTP connections.
Specify the endpoint URL and model ID. External URLs must use HTTPS. HTTP is limited to destinations on your local network. Because their data handling cannot be verified, do not send confidential information.
Lets you edit the instruction text used for minutes generation. It does not apply to on-device or local AI.
Fills in gaps from the recorded file. Processing takes longer.
Selects the microphone to use. Switching during a meeting can disrupt the recording.
Tap after connecting a Bluetooth device or similar.
Specifying it when known improves accuracy. With the built-in microphone, reverberation often merges speakers into one.
The threshold for judging silence. If it is too low, air-conditioning noise can prevent the silence timer from working.
Where minutes are saved automatically. The App creates its own folder inside the folder you choose. Exported files are never deleted automatically.
Only files inside the folder created by the App can be listed, renamed, and deleted. Deletion cannot be undone.
Exports the m4a audio together with the minutes. File sizes become large.
Lists the audio and minutes kept after minutes were created. Use "Play" to listen again, "Recreate minutes" to transcribe the audio again and rebuild the minutes (cloud AI usage fees apply if a cloud AI is selected), and "View minutes" to read the saved minutes.
Choose right after minutes are created, after 1 week, after 1 month, or never. The default is after 1 week. Minutes are never deleted automatically.
Available at any time.
Opens the contact form in your browser. Do not enter confidential information such as meeting audio, transcripts, minutes, or API keys.