Text where the caret is
Unlike Agent-only voice input, WhisperJot can type into a focused editor location, comment, commit message, or other writable Cursor control.
Cursor provides native voice input for the AI Agent prompt, activated by holding Ctrl+M in the Agents window, but it is not general editor dictation. For spoken text in code files, comments, commit messages, or other apps, use OS dictation or WhisperJot with the target field focused.
Cursor added native voice input with Cursor 2.0 and later updated it in the 3.1 changelog. It records while Ctrl+M is held in the Agents window and turns the full clip into an Agent prompt after release. This is a convenient way to talk to the Agent, but it does not provide general speech-to-text in a code file or arbitrary Cursor field. OS dictation and system-wide tools cover those other text-entry locations. Cursor has not documented where this transcription is processed in the cited materials.
Unlike Agent-only voice input, WhisperJot can type into a focused editor location, comment, commit message, or other writable Cursor control.
Teach custom vocabulary the names used by your codebase so uncommon packages, symbols, and product terms receive added context.
WhisperJot processes speech on-device by default with its local engines, which run fully offline following the roughly 2.3 GB model download.
A single activation shortcut follows you from Cursor into a terminal app, project notes, code review, or team conversation.
Yes, but it is scoped to the AI Agent prompt. In the Agents window, press and hold Ctrl+M while speaking, then release it. Cursor processes the full recorded clip into prompt text. This native feature does not act as general dictation inside code files or every arbitrary text field.
Cursor's cited voice feature targets the Agent prompt rather than the code editor. To dictate at an editor caret, focus the desired location and use operating-system dictation or a system-wide tool such as WhisperJot. Always inspect dictated code or commands carefully before saving or executing the result.
Cursor does not verify support for Microsoft's VS Code Speech extension in the cited materials, so this guide does not rely on that route. For editor-wide voice entry, use macOS Dictation, Windows voice typing, or a separate system-wide dictation application whose supported fields and operating systems are documented.
The Cursor 3.1 changelog describes holding Ctrl+M and batch-transcribing the complete clip, but the cited Cursor materials do not say where that speech transcription runs. Avoid assuming it is local or cloud-based. Consult Cursor's current documentation and privacy information if processing location affects your decision.
WhisperJot types into the Cursor field that has focus and does not require a Cursor extension. Its local engines process speech on-device by default and can work fully offline after the initial model download. Custom vocabulary can provide context for recurring project identifiers and technical terms.
Sources, checked August 24, 2026: Cursor 3.1 changelog, Cursor 2.0 announcement. Vendor features change — check each product's documentation for current behavior.
Browse WhisperJot use cases, compare dictation apps, or check the system requirements.
One $99/year plan includes unlimited dictation across your focused desktop apps.