Skip to main content
The Text Output settings control what HyperWhisper does when a recording stops. They set the automatic paste behavior, the cleanup of the text, and the clipboard behavior. Open the settings panel. Then select Text Output.
For more information about Auto Paste and how to correct faults, read Auto Paste & Clipboard Behavior.

Auto Paste

When Paste result automatically is on, HyperWhisper copies the transcript to the clipboard. It then sends a paste keystroke to the frontmost application. You do not change the focus or press a key.
Auto Paste needs the Accessibility permission. This permission lets HyperWhisper send the paste keystroke for you. On macOS, the home screen shows an accessibility prompt while the Accessibility permission is not given. The prompt shows also when Auto Paste is off. On Windows, Auto Paste needs no extra permission.

Give the Accessibility permission

  1. Go to System Settings → Privacy & Security → Accessibility.
  2. Unlock the list.
  3. Turn on the toggle next to HyperWhisper.
  4. If the app was open when you changed the setting, restart the app.

Remove Filler Words

Turn on Remove filler words to remove the speech fillers “uh”, “um”, and “er” from the raw transcript. HyperWhisper removes them before it delivers the text.
This setting applies only to a mode that does not use AI post-processing. When AI post-processing is on, the AI model cleans the transcript with the prompt of the mode. This toggle then has no more effect.
HyperWhisper removes filler words from English transcripts only, on macOS for both regular (mode-based) transcription and streaming, and on Windows during streaming. On Windows, regular (mode-based) transcription removes filler words regardless of the transcript language.

Autocapitalize Insert

Autocapitalize Insert changes the capitalization of the first word of your transcript. The new capitalization agrees with the position of the cursor in the target text field.
  • If the cursor is mid-sentence (regular text comes before it), HyperWhisper makes the first letter lowercase. The text then reads as part of the sentence.
  • If the cursor is at the start of a sentence, HyperWhisper does not change the transcript. This position comes after a sentence-end character such as ., !, or ?, or at the start of a field.
  • Acronyms stay unchanged — if the first word starts with two or more uppercase letters (for example, “API” or “USA”), HyperWhisper never makes it lowercase.
  • macOS keeps the first-person pronoun “I” — macOS never makes “I”, “I’m”, “I’ll”, “I’ve”, and “I’d” lowercase, also mid-sentence. This includes the forms with a curly apostrophe. Windows does not have this guard yet, and it can make a mid-sentence “I” lowercase.
Autocapitalize Insert reads the cursor position with the Accessibility API. When you turn on the toggle the first time, macOS asks for the Accessibility permission. If you cancel the prompt, the toggle goes back to off.If you want the dictated text exactly as transcribed at all cursor positions, turn off this setting.

Spacing Between Transcriptions

HyperWhisper adds one space to the end of each transcript before it delivers the text. The space keeps two recordings apart when you dictate one after the other. Without it, the two transcripts run together as “Hello world.How are you?”. This behavior is automatic. There is no setting for it. HyperWhisper does not add the space in these three cases:
  • The transcript already ends with a space or a line break.
  • The transcript is empty.
  • The language of the mode is a language that writes no spaces between words.
The languages that write no spaces are Japanese (ja), Chinese (zh, and the zh-TW, zh-Hans, and zh-Hant variants), Cantonese (yue), Korean (ko), and Thai (th). A regional variant such as zh-CN counts as Chinese. When the language of the mode is Automatic, HyperWhisper reads the transcript instead. If more than 30% of the characters are CJK characters — the punctuation and the spaces are not counted — HyperWhisper adds no space.

During streaming

Streaming delivers the transcript in parts, so the rule applies to each part.
  • On macOS and Windows, HyperWhisper adds the trailing space to each part, with the rule above. The streaming language decides, and not the mode language.
  • On Linux, HyperWhisper puts a separator between the parts instead. The separator is empty for a no-space language, and one space for the other languages. When the streaming language is Automatic, Linux examines the text before the boundary and the next part. If one of them is written in a continuous script — the CJK scripts or Thai — the separator is empty.

Clipboard Settings

The Clipboard section controls the clipboard behavior before and after Auto Paste.

Restore clipboard after paste

When this setting is on, HyperWhisper saves the contents of your clipboard before it pastes the transcript. It then restores the contents. A transcript does not permanently replace the content that you copied before.

Restore delay

When the clipboard restore setting is on, a delay control shows. You can set the restore delay from 1 to 60 seconds. A long delay gives slow applications sufficient time to complete the paste. HyperWhisper then replaces the clipboard content with your previous content.
Use the stepper control to increase or decrease the delay in steps of one second.

Hide from clipboard history

When Hide from clipboard history is on, HyperWhisper writes the transcript to the clipboard with a marker. The marker tells clipboard manager applications not to save the text. Use this setting when you dictate sensitive content, such as passwords, private notes, and confidential messages. These entries do not show in tools such as Paste, Maccy, Alfred, or Windows Clipboard History.
HyperWhisper puts the transcript on the clipboard with a type marker. This marker tells the clipboard managers that support the convention (Paste, Maccy, Alfred, and more) not to save the transcript.
The exclusion works only if the clipboard manager supports the convention. A manager that ignores the exclusion flag saves the entry.