How do I directly type into an app's text field?

[justauserid]

For example I am trying to type directly into a TextEdit opened file, but the text is not being typed directly in the file but first on app overlay and then when click the hotkey it goes to the TextEdit file. What settings need to be changed so that voice input directly types text into an editable text field on the mac?

[MLuc24]

What you are seeing is the intended flow rather than a setting that is switched the wrong way, so it is worth knowing why before hunting for a toggle.

FluidVoice does not type into the target app as you speak. It captures where your text should go before it shows anything, then puts the finished text there in one go at the end. The comment in ContentView.swift is explicit about the ordering:

private func captureRecordingTargetContext() {
    // Capture the focused target PID BEFORE any overlay/UI changes.
    // Used to restore focus when the user interacts with overlay dropdowns.
    let focusTarget = TypingService.captureSystemFocusTarget()

That capture exists because the overlay necessarily steals focus while you dictate. TypingService then delivers the result by putting it on the pasteboard and simulating Cmd+V into the element it remembered, restoring your previous clipboard afterwards. It is a capture-then-insert design, not a live-typing one, so there is no setting that turns it into keystrokes arriving in TextEdit while you talk — macOS's built-in Dictation is the thing that behaves that way.

The text appearing in the overlay first is a preview of the transcription in progress, controlled by enableStreamingPreview. Turning that off stops the words showing up there as you speak, so nothing appears until the text lands in TextEdit — which may be closer to what you want visually, though it does not change when the insertion happens.

One thing worth separating from the design question: if the second hotkey press feels like an extra step rather than "stop recording", or if text sometimes lands in the wrong app, check that FluidVoice has Accessibility permission in System Settings → Privacy & Security. Both the focus capture and the synthetic Cmd+V depend on it, and when it is missing or was granted to an older copy of the app the insertion half degrades while the recording half keeps working — which looks exactly like text getting stuck in the overlay.

[justauserid]

Accessibility permission is granted. If not per word streaming, I at least wanted a chunked streaming and automatically (w/o having to press a key for that). I am also always fearing the data loss (irrespective of the stated history feature). If the text is in my app of choice then data loss is my worry beyond that.

But the main concern is it just feels not intuitive. It doesn't feel like I am writing. Had it been for Claude prompts etc I would not even have been looking for a tool and just rocked the Apple's offering at whatever level of quality and speed it does it.

Is there any other app/tool (you might know of) that does it easily on a mac with local models?

Thanks for the detailed reply.

[altic-dev]

Yeah, typing it directly would cause problems in fixing your mistakes, which happens after that. So the right way is to show what you talk in the box first and then it goes into the field once you're done.

[justauserid]

Added a bit in another comment why I was looking for it: #896 (reply in thread)

But yeah I get this point. Just thought it will be nice to be somewhere in the middle and at least sending/streaming last full sentence back to text editor field, or at least a choice to do that.