When something doesn't work

Is the cursor in a text field? Without a text field, chatFlow deliberately pastes nothing and shows “No text field — the text is in the buffer.” Then click into the field and press the buffer hotkey (default: ⌘ right).

If chatFlow doesn't have the “Accessibility” permission, it can't paste. The text is then on the clipboard, and you paste it yourself with ⌘V. To enable it: System Settings → Privacy & Security → Accessibility → turn on chatFlow. As long as the permission is missing, a notice appears at the top of the menu bar menu.

In the main window, check under General → Record Hotkey which key is selected, and whether chatFlow is allowed under Accessibility. Don't use a plain ⌘ key: pressing ⌘ twice starts macOS dictation.

That's macOS dictation, not chatFlow: pressing ⌘ twice starts it. Turn it off under System Settings → Keyboard → Dictation, or change the shortcut there. Or choose a different record key in chatFlow, for example the right Option key.

The key was released too early. Finish speaking and let go a moment later. With Flow Max, chatFlow keeps recording briefly after you let go; with the local tiers, the recording ends exactly when you let go.

Does chatFlow have the “Microphone” permission? System Settings → Privacy & Security → Microphone → turn on chatFlow.

If the microphone is set too quiet or too loud, recognition suffers. In the main window, under General → Input volume, Check level measures for five seconds while you speak and suggests a better value if needed.

Under General → Recognition Language, set a fixed language instead of “Auto-detect”. With Flow Air the option is grayed out; choose Flow Pro or Flow Max there first. Automatic detection can get it wrong with short or mixed sentences.

Too many: lower the frequency in percent or the maximum per sentence. None at all: check whether the trigger word really appears in the text and the emoji is turned on. With conflict pairs you prevent two emojis from appearing together.

Tell Flo what should be different. He'll switch the matching text rule.

Check in this order: hotkey pressed (the pointer turns into a crosshair), trigger word plus request spoken meanwhile, an area selected? Does chatFlow have the “Screen Recording” permission (System Settings → Privacy & Security)? Without this permission, the image command doesn't start at all. It also needs a sign-in and enough credit. Each model has its own trigger word; if the field is empty, the model is off.

It's almost always the material, not the tier. In this order:

1. Run Check clips and only use the clips marked green (the Best button selects exactly those).
2. Untick the clips that failed.
3. Process: free AI denoise; for music or noise in the background Separation · Pro (20 credits per started minute); for reverb Remove reverb.
4. Record more clean material: a quiet room, little reverb, the same distance from the microphone.
5. Only then a higher tier or a trained model.

To the right of it, Diagnostics opens a window with the full error text and the state of the background engine. With copy all you can send it to us.

The local tiers keep working: recognition, text rules and local translation, with no credits. Everything that goes through the chatFlow cloud needs internet: the cloud tiers, the AI voice, the image command, audio processing in the studio, the speech analysis, voice training, Spark, polishing, Agency, chat and Flo. Your subscription status is also checked online.

Usually the language model is briefly overloaded, and chatFlow then switches on its own. The dots show it's still working; wait a moment or ask the question again.