Updated October 8, 2026 (for version 2.1)
The app starts in Simple Mode. Tap Choose a file or Record, and choose the number of speakers. Transcription and speaker identification run on your device. No sign-in or API key is needed.
Recording continues while the screen is locked. Calls or system limits can interrupt it. There is no limit on recording length. If the app quits while recording, it tries to recover the recording the next time you open it (recovery may not always succeed).
Download files stored in iCloud Drive or other cloud services to your device first. Any audio format your iPhone can play is supported. If a file cannot be imported, check that it plays in its original app, export it in another format, and try again.
If Prepare model to continue appears, follow the instructions to prepare the model first.
If the file cannot be accepted while recording, processing, editing, selecting a file, or viewing another screen, finish that task and share the file again. Follow any message about an unsupported format or insufficient storage. The original audio file in the sharing app is kept.
On iOS 26 or later, transcription can continue on your device when you switch to another app. iOS controls how long it can continue, so it may stop before finishing. If iOS cancels continued processing or its time expires, processing stops and does not restart automatically. On earlier versions of iOS, keep the app in the foreground until transcription finishes.
After an interruption, open the app and choose Continue if a resume prompt appears for the same audio and settings. Processing continues after the saved chunks; the chunk that was in progress is redone. You can also choose Start over. If there is no saved progress, processing starts from the beginning. Your original audio is kept. See resuming a transcription for the conditions.
Tap Detailed Mode at the top right of the home screen, or Switch to detailed mode at the end of the Simple Mode settings. To go back, tap Simple Mode at the top right, or Return to Simple Mode in Settings.
Simple Mode does not use Detailed Mode settings such as method, quality, or number of speakers. Those settings are kept and are ready to use when you return to Detailed Mode. The speaker choice in Simple Mode is saved separately, too. Switching modes or reopening the app does not replace one mode's settings with the other's. For example, Simple Mode can keep 2 people while Detailed Mode keeps 9 people.
If there are no saved Detailed Mode settings yet, your Simple Mode choices provide the starting point only the first time you switch. The save location, retention period, app lock, dictionaries, history, and display language are shared by both modes.
Speech recognition can make mistakes. Check important numbers, names, and negations against the original audio. When a recording has heavy background noise, the app lets you know that accuracy may drop.
When you choose to share or save to a cloud service, the text is sent to the app or service you chose.
Speaker identification runs on your device. By default, two methods check each other, and parts where they disagree are marked Needs review. In Detailed Mode, Speaker identification in Advanced Settings lets you use only one of them. Similar voices or people talking at the same time can be mixed up, so listen again to check. To change speaker names, choose Rename Speakers… from Other. To try again with a different number of speakers, choose Other › Transcribe again.
Speakers can be identified in audio up to 60 minutes long. For longer audio, transcribe with 1 person, or split the audio into parts of 60 minutes or less before importing it.
Whisper can transcribe audio up to 4 hours long. For longer audio, split it into parts of 4 hours or less, or choose iOS built-in.
Passages such as removed stock phrases are automatically re-listened to on your device, and suggestions are shown if available. Separately, you can enable additional re-listening in Detailed Mode under Settings › Advanced Settings › Recognition and text options › Re-recognize unclear segments (off by default). Re-listened results can also contain errors, so check them against the original audio. To restore the original settings, use Reset Transcription Settings in the Detailed Mode settings. Recordings, history, dictionaries, the retention period, and the app lock are kept.
In Simple Mode, open Settings › Data › Register Names and Terms (Dictionary). In Detailed Mode, open Advanced Settings › Dictionary. Enter the correct spelling and, if needed, the reading and spellings that are often misrecognized. You can create several dictionaries and import many words at once from a CSV file (in the form “spelling,reading”). Dictionaries are stored on your device. Registering a word does not guarantee that it will always be recognized correctly.
You can buy an optional tip from Settings › About This App › Support Development. Purchases are handled by Apple's App Store. Buying a tip does not change the available features, and choosing not to buy one does not restrict any features.
In both Simple and Detailed Mode, a card may appear below the text just after a transcription succeeds in the foreground and its result is shown. It appears only when all these conditions are met:
Days are counted in 24-hour periods. Successful transcriptions are counted from the introduction of this feature. If no first-use date is saved, the first launch after updating is used. The card does not appear just after launch, during recording, transcription, or background processing, while another sheet or confirmation is open, or while the app is locked. It does not appear for failed, cancelled, or partial results. A missed card is not shown later when you close another screen, return to the app, or open History.
Choose Support development to open the tip screen, or Not now to close the card. Once closed, the card does not reappear for the same result. The interval and success count for the next appearance are counted from when the card was shown. To stop future cards, choose Don’t show again, or turn off Show tip requests in Settings › About This App (it is on by default). Turning this setting back on allows cards again even after you choose Don’t show again.
The dates of use, card appearances, and purchases, successful transcription counts, and display preferences used for the card are recorded only on your device and are not sent out.
Separately from the tip card, the system rating prompt may appear after a successful transcription. The app asks the system at most once per version on this device. The system decides whether to show the prompt, so it may not appear.
This requires at least 5 successful transcriptions and at least 7 days since first use. The transcription must succeed after the app has been in the foreground for at least 30 seconds; the app checks the conditions again 2 seconds after displaying the result. It skips the request while the tip card is visible, on the same day as a card appearance or the latest purchase, and within 24 hours of either. It also skips it during recording or processing, while another screen is open, or while the app is locked. Rating requests are separate from tip preferences and whether you have bought a tip: turning off Show tip requests does not disable them. To rate the app yourself, open Settings › About This App › Rate on the App Store. The tip card has no rating button.
Information that helps investigate the problem is attached automatically, such as device and app information, the processing stage, and error codes. Crash, hang, and other diagnostic information received from Apple's MetricKit is included when available. Automatically attached information does not contain transcription text, audio, or API keys. Opening the report screen, or tapping Send to open the email composer or share sheet, does not by itself send anything to the developer.
You can optionally choose one screenshot from Photos. Image metadata such as location is removed before attachment, but text and personal information visible in the image are not removed. Check the image before sending it. If you choose a record containing transcription text under Details, it is included only after you review it and confirm the attachment. Records containing transcription text are off by default and are not attached automatically.
Reports you confirm sending are received at the developer's email address. See Support for contact details.
Yes. Transcription and speaker identification run on your device, and audio and text are not sent to outside services for them. Network access may be used for model or recognition-resource downloads, files stored in the cloud, sharing, Writing Tools, bug reports, tip purchases, and App Store ratings. See the Privacy Policy for details.
Transcription in this version has no additional charge. Tips to support development are optional and do not change any features. Data charges for downloads and sharing depend on your plan.
Yes, in Detailed Mode on supported devices (iOS 26 or later). In Simple Mode, the text appears after recording.
Recording continues while the screen is locked, but calls or system limits can interrupt it. On iOS 26 or later, transcription after recording can continue when you switch to another app. If continued processing is cancelled or expires, transcription stops immediately and does not restart automatically. The next time you open the app, you can choose Continue if saved progress is available. On earlier versions of iOS, keep the app in the foreground until transcription finishes.
Yes. With Whisper, audio longer than 20 minutes is processed in chunks of about 12 minutes, and finished chunks are saved. If transcription stops and a resume prompt appears for the same audio and settings, choose Continue (you can also choose Start over). The chunk that was in progress is redone. Short audio is not saved until recognition finishes. iOS built-in starts from the beginning. Partial results are kept for 48 hours by default. You can change this in Detailed Mode under Settings › Advanced Settings › Recognition and text options › Retention period for partial results. Transcription may also start from the beginning if this period has passed or there was not enough storage to save. Your original audio is kept.
Long audio and the Highest quality take longer. Check the stage, elapsed time, and progress on screen, and wait. Processing may also take longer than usual when your device is hot.
Choose one file at a time from the Files app. Any audio format your iPhone can play is supported. Whisper can transcribe audio up to 4 hours long, and speakers can be identified in audio up to 60 minutes long.
Check that the app is allowed to use the microphone. If you did not allow it, change it from Open Microphone Settings on the recording screen. If a message about storage space or the save location appears, follow it.
They may be recognition errors. They can also be real English words or names, so the app does not delete them automatically. Stock phrases (such as “thank you for watching”) are removed from the text when silence, low confidence, or similar evidence suggests a recognition error, and the part is marked “(Could not make out m:ss–m:ss)”. Removed text can be restored from Needs review. Listen to the original audio, for example from Needs review, and edit the text if needed.
Display language in Settings opens the iOS language setting for this app. The display language and the language of speech recognition are separate. Speech processing is designed for Japanese.
In Simple Mode, use Settings (gear) › Text size. In Detailed Mode, use Advanced Settings › Display and storage.
Use Report on a message, Other › Report a problem on the results screen, or Bug reports & requests in Settings. Automatically attached information does not include transcription text or audio. Screenshots are optional. See Report a problem for the steps.