Short answer: Jira has no dictation feature of its own, so failures come from macOS or your browser. Check the browser's microphone permission in System Settings, click inside the description or comment box so the rich text editor holds focus, disable any password manager causing Secure Input, and watch for the dictation timeout on long tickets.
You open a Jira ticket, click into the description field, hit your dictation shortcut, and start describing the bug. The microphone indicator lights up. You talk for thirty seconds. The description field is still empty.
Or a variation: the first sentence lands perfectly, then the text stops arriving halfway through the second one. Or you say "slash" somewhere in a sentence and the whole editor suddenly pops open a command menu. Or you dictate a summary, tab away, and Jira saves an empty field over the one you just filled in.
Jira is one of the harder places to diagnose dictation problems, and not because Jira is broken. It is because Jira has no dictation feature at all. There is no microphone button in the issue editor, no voice setting in your profile, nothing in project settings to toggle. Every part of the chain that can fail sits either in macOS or in the browser between macOS and Jira. That narrows the problem considerably, and it means the fixes below are testable in order.
Work down the list and stop when your words land in the field.
First: a 30-second test that tells you where the problem is
Before touching any settings, find out whether dictation is broken everywhere or only in Jira. Open TextEdit, create a new blank document, and dictate a sentence.
- Text appears in TextEdit: macOS dictation is healthy. The problem is your browser, the Jira editor, or a focus issue. Skip to Fix 2.
- Nothing appears in TextEdit either: the problem is system-wide. Start at Fix 1 and work through Fixes 6 to 8.
This single test saves most people ten minutes of poking at the wrong settings. Almost every "it only breaks in Jira" report turns out to be a browser permission or a focus problem, and almost every "it broke everywhere at once" report turns out to be an input device or a permission that got reset by an update.
Fix 1: Give your browser microphone access in System Settings
macOS gates microphone access per application. If you use Jira in Chrome, Chrome needs microphone permission. If you switched from Chrome to Safari last month, Safari has its own separate grant and it may never have been given.
Go to System Settings → Privacy & Security → Microphone and confirm the toggle next to your browser is on. If it is already on, switch it off, wait a few seconds, and switch it back on. That forces macOS to reissue the permission, which resolves a surprising number of cases where the toggle looks correct but the grant has gone stale.
Quit the browser completely after changing this. Closing the window is not enough; the permission is read when the process starts. Use Cmd+Q, confirm the browser is gone from the Dock, then reopen it and load Jira again.
macOS point updates are the usual reason this breaks without warning. Permissions that were granted under an older version sometimes need to be reconfirmed, and nothing tells you it happened. If your dictation stopped the same week you installed a system update, this is almost certainly the cause.
Fix 2: Check the site-level microphone permission inside the browser
There are two permission layers, and people routinely fix one while the other stays blocked. macOS decides whether the browser can use the microphone at all. The browser then decides, per website, whether that specific site can.
Click the icon at the left edge of the address bar while you are on your Jira site. You will see a microphone entry with Allow, Block, or Ask. If it says Block, change it to Allow and reload the page.
This matters more than it sounds for Jira specifically, because most companies run Jira on a custom subdomain. Your permission for company.atlassian.net is separate from any permission for other Atlassian properties, and separate again if your organization moved to a different domain during a migration. A site permission that was set years ago on the old URL does nothing on the new one.
Fix 3: Make sure the rich text editor actually has focus
This is the single most common Jira-specific cause. The Jira issue editor is not a plain text box. It is a rich text editor built out of nested elements, and it has a real edit mode it has to enter before it will accept keystrokes.
If you click near the description field but not precisely inside the editable area, the page looks like it is ready for input while focus is still on the page body. Dictated text has nowhere to go, so it goes nowhere.
The fix is deliberate: click directly on the text area until you see the blinking cursor, type one character with the keyboard to confirm it registers, delete it, then start dictating. If that single character does not appear, dictation was never going to work either. You have proven the problem is focus, not the microphone.
The same thing applies to comment boxes. Jira comment fields on an issue page often start as a collapsed placeholder that says something like "Add a comment." That placeholder is a button, not a text field. You have to click it once to expand the real editor, and only then does the editor accept input.
Fix 4: Watch out for slash commands and @ mentions
The Jira editor treats certain characters as triggers. Typing a slash at the start of a line opens the insert menu. Typing an at sign opens the mention picker. Typing a colon can open the emoji picker.
Dictation produces those characters when you say their names. If you dictate "the fix is in the slash API slash users endpoint," the editor may interpret the first slash as a command trigger and open a floating menu, which then steals focus from the text you are still speaking. The rest of the sentence disappears into the menu's filter box.
Two workarounds. First, when you know a field will contain slashes or at signs, dictate the prose first and add those characters by hand afterward. Second, avoid starting a line with a slash. The trigger is far more aggressive at the beginning of a line than mid-sentence, so rephrasing so the path is not the first thing on the line usually avoids it entirely.
If you write a lot of tickets containing file paths and endpoints, our guide to dictating quotes and special characters covers the commands for punctuation that speech engines handle reliably.
Fix 5: The dictation timeout is cutting your long descriptions off
Built-in macOS dictation is not designed for long continuous speech. It listens, transcribes, and then stops on its own after a stretch of continuous input or after a pause it reads as the end of your sentence. You keep talking. It stopped listening a while ago.
This is why the failure so often looks like "the first part worked." A bug description is naturally long. You set the scene, list reproduction steps, describe expected versus actual behavior. By step three you are well past where the built-in dictation session ended.
The workaround with built-in dictation is to dictate in short bursts, one or two sentences at a time, restarting the session between each. It works, but it turns a two-minute ticket into a five-minute one and it breaks your train of thought exactly when you are trying to describe something precisely. We covered the mechanics of this in detail in our piece on the Mac dictation 30-second limit.
Fix 6: Secure Input is silently blocking keystrokes
macOS has a mode called Secure Input that prevents applications from reading keyboard events. It exists so password fields cannot be logged. When it is active, dictation cannot deliver text.
Password managers are the usual trigger. So are terminal windows waiting on a password prompt, locked note fields, and any app that opened a secure field and did not properly release the lock when it closed.
The frustrating part is that nothing tells you Secure Input is on. There is no menu bar icon, no notification. Text just stops arriving.
To clear it: close any open password manager window, close terminal windows sitting at a password prompt, then click on a plain text field in a different app and try dictating there. If that works, the culprit was whatever you just closed. If it persists, log out of your macOS account and log back in. That always clears it.
Jira users hit this more than most because of how a workday looks. You unlock a password manager to get into a company SSO portal, the portal hands you off to Jira, and the password manager window is still open behind the browser holding the lock.
Fix 7: Your Bluetooth headset is in the wrong audio mode
Bluetooth headsets have two profiles. One is high-quality stereo output with no microphone. The other enables the microphone but drops audio quality noticeably, because the connection has to carry audio in both directions.
macOS switches between them based on what it thinks you need, and it does not always guess correctly. If your headset stays in output-only mode, the microphone is not available even though the headset is connected and you can hear audio through it perfectly.
Test it by opening System Settings → Sound → Input and speaking. If the input level meter does not move, the microphone is not live. Switch the input source to your Mac's built-in microphone and try dictating again. If it works, the headset was the problem.
This bites hardest for people who write tickets right after a standup. You finish the call, the headset is still connected in call mode or has half-switched back, and the first ticket you try to dictate goes nowhere. Disconnecting and reconnecting the headset usually resets it. Our post on Mac dictation not working with AirPods goes deeper on this specific pattern.
Fix 8: The wrong input device is selected
Related but distinct. Your Mac may be listening to a device that is not near your mouth: an external monitor's built-in microphone, a webcam, a desk dock, a virtual audio device installed by a screen recorder or conferencing app.
Open System Settings → Sound → Input, look at what is selected, and watch the level meter while you talk normally. If the bars barely move, the selected device is not picking you up. Choose the correct one.
Virtual audio devices deserve special attention. Screen recording tools and meeting apps install them, and they sometimes set themselves as the default input on launch and do not restore the previous setting when they quit. If you were on a call before your dictation stopped working, check this first.
Fix 9: Inline edit fields are saving before you finish
The summary field on a Jira issue, and several fields in the right-hand details panel, use inline editing. You click, the field becomes editable, and it commits when it loses focus or when you press Enter.
Dictation interacts badly with this. If the dictation session ends and something else takes focus while you are still mid-thought, Jira commits whatever partial text exists at that moment. You look up and the summary reads "Login button does not" because that is where the session stopped.
Worse, if the field was committed while empty, you may have overwritten a summary that was already there. Jira keeps a history on the issue, so the previous value is recoverable, but it is an unpleasant surprise.
The safe pattern for inline fields: dictate into a plain text scratch space first, read it back, then paste it in. Reserve direct dictation for the description and comment fields, which do not auto-commit the same way.
Fix 10: Check your dictation language and your vocabulary
If text is arriving but it is wrong, this is a different problem with a different fix. Two things cause it.
The first is a language mismatch. Check System Settings → Keyboard → Dictation and confirm the language matches what you are speaking, including the regional variant. A US English speaker with dictation set to UK English will see consistent small errors that look like random inaccuracy.
The second is vocabulary, and this is where ticket writing is genuinely harder than ordinary prose. A Jira description is dense with terms no general speech model expects: service names, repository names, internal acronyms, library names, error codes, ticket keys. Generic dictation renders these phonetically. Your service called "Kestrel" becomes "chestral." Your ticket key becomes a spelled-out mess.
Built-in macOS dictation gives you no way to teach it these words. That limitation is structural, not a setting you have missed.
Why ticket writing breaks dictation harder than normal writing
Step back and the pattern is clear. Everything that makes a Jira ticket useful is exactly what general-purpose dictation handles worst.
Tickets are dense with proper nouns. They contain code identifiers and file paths. They are structured, with numbered reproduction steps and headings. They mix prose with literal strings that must be exact. And they are long enough to outrun a dictation session that was designed for a text message.
That is why "fix the permission and it works" is only half the story for Jira. Once the microphone is working you hit the second wall: the transcription itself is not tuned for the words you actually use at work.
A different approach to dictating tickets
The fixes above address the built-in path, and for many people one of them solves it. If you write tickets all day, though, there is a structurally different option.
Voice Keyboard Pro is a Mac app that lives in the menu bar. You hold a hotkey, speak, release, and the text appears at your cursor in whatever app you are already in. That includes the Jira description field, the comment box, and anywhere else in the browser. There is no separate window, no copying across, no dictation panel to manage.
Three differences matter for Jira work specifically.
The first is Smart Vocabulary. You add your service names, repository names, internal acronyms, and project keys to a personal dictionary with replacement rules, and Voice Keyboard Pro's transcription engine applies them every time. Your team's jargon stops coming out phonetically. This is the difference between a ticket you can submit and a ticket you have to spend two minutes correcting.
The second is that there is no short session cutoff to work around. You hold the key for as long as you are speaking. A full reproduction sequence with five steps comes out as one continuous piece of text, so you can describe a bug the way you would explain it to a colleague instead of rationing it into fragments.
The third is that it works the same way in every app. The same hotkey that fills a Jira description fills a Slack message, a pull request body, or a document. Once the gesture is in your hands it does not change per app, which is the opposite of the current situation where every tool has its own quirks. If your work spans Atlassian tools, the same applies in Confluence.
On privacy: the Voice Keyboard Pro server stores only operational pings. No audio and no transcript content is retained, which matters when the text you are dictating is an internal bug report describing how a system fails.
There is a free tier with daily limits. Pro is $4.99 a month or $34.99 a year.
Quick reference
- Nothing appears anywhere: browser microphone permission in System Settings, then input device.
- Works in TextEdit, not in Jira: site-level browser permission, then editor focus.
- First sentence lands, rest disappears: dictation session timeout.
- Text stops when a menu pops up: a slash or at sign triggered the editor's command menu.
- Stopped working today with no changes: Secure Input from a password manager, or a macOS update that reset permissions.
- Summary field keeps saving half a sentence: inline edit committing on focus loss. Dictate elsewhere and paste.
- Text arrives but the technical terms are wrong: vocabulary, not permissions. Built-in dictation cannot be taught your terms.
Ten steps looks like a lot, but the test at the top tells you which branch to take, and the first three fixes resolve the large majority of cases. Most people are back to filing tickets inside two minutes. If you want to go further and structure the whole ticket by voice, our guide on dictating bug reports covers what to say and in what order.
The words land at your cursor in whatever app you are already in — Mail, Slack, Word, a browser form. No dictation window, nothing to copy across.
Free forever for casual use · Apple Silicon & Intel