Changelog

✦ Fixed

• The app can no longer be frozen by writing a log line — when VoxiPrompter is started from a terminal or from another program, its diagnostic output goes to whatever started it. If that reader stops accepting text, every attempt to write a line used to wait forever, and one waiting line was enough to stall everything else: slides stopped moving, the phone page stopped answering, and the app sat at zero percent processor looking perfectly healthy. Diagnostic output is now handed off and never waited on, and app.log still records every line
• If anything else ever stalls it, the log now says so — a new line reading Runtime stall names how long the app was unresponsive, instead of leaving the cause to be guessed at afterwards

✦ Connecting to OBS

• The port and password fill themselves in — OBS writes them to a file on this computer, so the first time you open the OBS panel they are already correct. A Detect button re-reads them whenever you want, and says what it changed rather than changing it behind you
• “It will not connect” now tells you why — the most common cause is that OBS’s WebSocket server is simply switched off, which used to look identical to a wrong port or a firewall. The app now says which one it is, and where the switch lives
• Nothing is changed behind your back — if you typed a host yourself, or OBS is on another computer, the app leaves it alone and only mentions the disagreement

✦ Slides and overlays that are not listening

• You can now see what is connected — every slide page, overlay or other tool attached to VoxiPrompter is listed by name, with when it last received a cue. A source that is pointed at the wrong address is simply absent from that list, which is the whole point: it used to be indistinguishable from one quietly waiting
• The address to give them is shown, with a copy button — and if VoxiPrompter had to move to a different port because something else was using its usual one, it says so in red, because anything you saved earlier is now pointing at nothing

✦ For anything you connect to VoxiPrompter

• Rehearsal mode is now announced properly — it has been in the app since v0.9.1, but a connected tool asking whether it was available was told no. A setup that hid its rehearsal features against a version that had them will find them now
• Every capability the app reports is now checked against its own documentation — the list a connected tool reads and the list in our integration guide can no longer drift apart, because a mismatch stops the release
• We now say plainly which changes reach you whether you ask for them or not — some improvements only happen when a tool opts in, and some change what every connected tool receives on the next take. Those two are labelled differently now, and the second kind has to spell out what your setup will do differently before it ships

✦ Under the hood

• Scene and source cues are now tested against a stand-in for OBS — scene changes, showing and hiding a source, and timed cues are checked on a real connection before every release, including what happens when you step back through a line that carries more than one cue
• Two tests that could never fail were rewritten — they were watching the highlight behave and quietly accepting any answer

✦ Fixed

• Clearing the script and pressing Start no longer erases it — the confirmation you were meant to see was being skipped by that exact sequence. It now always asks first
• Editing on your phone no longer overwrites what you are typing on the desktop — and the cursor no longer jumps to the end of the box mid-sentence
• The highlight stays with you in long paragraphs — on a script pasted as one paragraph per line, a phrase you repeat could throw the highlight to the far end of that paragraph. It now stays inside the search window you set
• Slide cues written on the same line both fire when you go back — a heading carrying a slide cue and a scene change used to send only the scene when you stepped back a paragraph
• Brackets written without spaces around them line up correctly — one of those anywhere in a script used to make every cue after it fire a word late
• The second screen stops keeping your monitor awake — after a session it went on holding the screen on until the window was closed
• Pro no longer disappears when your Mac changes network name — joining a different Wi-Fi could silently drop you to the free limits. Existing activations carry over with nothing to re-enter
• Disconnecting from OBS mid-connection stays disconnected — pressing Disconnect while a connection test was still running could leave a live connection behind a screen that said otherwise

✦ Faster and quieter

• Cleaner listening on built-in microphones — high-pitched noise from a fan or a power supply used to be folded down into the range of your voice. It is filtered out before recognition now
• Long scripts track your voice up to a hundred times faster — a script pasted as one long block used to cost far more work per spoken word than the same text with line breaks. Both are quick now
• Far less disk activity during a take — the app was saving your position to disk on every recognised word
• Smoother teleprompter on every screen — the display, the phone and the OBS overlays were redrawing styling on every update instead of only when something changed

Nothing about how you write scripts or run a take changes in this release — everything above is the same features, working the way they were meant to.

✦ What's New

• Only your words on the reading screens — the brackets addressed to the app, such as scene changes, slide cues and sound effects, are no longer drawn on Desktop Display, Mobile Display or the overlay teleprompter. On a real 1,600-word script that is 30% less to scroll through, and a phone screen shows 23 words at a time instead of 16
• The pause direction stays — [JEDA] is the one bracket you perform rather than the app, so it is still on screen. Hiding it would remove direction, not noise
• Cleaner on camera — the overlay teleprompter is a browser source in OBS, so a bracket left on screen is one that can end up composited into the recording. Those are gone now
• Live Monitor is unchanged — it still shows every marker, which is where you go to check them

✦ Fixed

• The first status message no longer goes missing — on a display screen the opening status line, and the first check that keeps the screen awake, were both being swallowed the moment the connection came up

Your place in the script is tracked exactly as before — this changes only what is drawn, never where the highlight lands.

✦ What's New

• Rehearsal mode — walk the script on a clock with the microphone off. Scenes switch and slides move exactly as they will on the day, because it drives the same machinery a spoken read does. You can start from any cue, and it will put OBS and your deck into the state that point needs first
• The script can start and stop the recording — write [REC: start] and [REC: stop] and OBS records between them. The real win is that the recording's zero point is known, so chapter and caption timestamps are exact instead of hand-corrected after you trim the head. A backward jump never stops a recording, so going back for a retake cannot end the take you are rescuing
• Broken brackets are reported when you load the script — with a line number, instead of being dropped in silence and discovered mid-take as a slide that never moved. A missing closing bracket is the one worth having: it is not dropped at all, so the presenter reads it aloud
• Cues can fire a few words early — there is a natural lag between saying a word and the app hearing it, measured at one and a half to two words. You can now ask cues to fire that many words ahead, instead of writing markers a beat early and hoping
• Better tracking when you paraphrase — if you reorder a phrase or slip an extra word in, the highlight stays on the word you are actually saying. Measured on real scripts, accuracy on paraphrased delivery went from around 70% to around 90%, with no loss in how well it ignores genuinely off-script talk
• Sources you toggle by hand are put back — hide a source manually, re-read the passage, and the marker now restores it. Before, it was discarded as a repeat. Scene markers already worked this way
• OBS failures reach you — a dropped connection, a timeout or a failed sign-in used to be completely silent, indistinguishable from a marker that was never written. All of them are reported now, naming the marker that was lost

✦ Fixed

• The connection port stopped wandering between launches — the shipped default sat inside the range macOS hands out to any program that opens a connection, so something unrelated could be holding it when the app started. The default is now 8765, and 8790 for mobile
• Two windows saving settings at once — one of them could report a failure even though the settings on disk were fine

Shipped to you as part of v0.9.1. Everything here keeps its earlier behaviour by default, so a connected tool written against v0.8.28 sees no change until it asks for one. Rehearsal mode, the recording markers and the cue lead all came from a beta tester running VoxiPrompter in YouTube production.

✦ What's New

• Typos in scene names are caught before you record — when you load a script, VoxiPrompter checks the scene and source names against OBS and warns about any it cannot find. Previously a misspelled name only showed up mid-take, as a shot that never changed
• Your script corrects the scene again — if you switch scene by hand during a take, the next marker naming that scene now puts it back. Before, VoxiPrompter compared against what it last sent rather than what OBS is actually showing, so a marker repeating the live scene did nothing at all
• Connected tools know which script they are following — the app now stamps each cue with a short fingerprint of the script, so a slide deck built from an older revision can tell instead of silently showing the wrong slides for a whole recording
• Cleaner slide changes — when several markers are crossed at once, connected tools are told which one is the last, so a deck can jump straight there instead of flicking through the ones in between

✦ For connected tools

• Slides can match your scene timing — if a marker waits before switching a scene, connected pages are now told how long that wait is, so a deck can move in step with the camera instead of guessing
• Tools recognise which VoxiPrompter they are talking to — the app now introduces itself with its version and what it can do, so an integration adapts itself instead of assuming

Nothing changes for the teleprompter itself. Both come from a beta tester running VoxiPrompter in YouTube production.

✦ Improvements

• Better timing for markers at the end of a line — a marker closing a line, or sitting on its own line, now fires on the last word before it. The scene is ready as you finish the sentence, instead of switching once you are already a word or two into the next one
• Pausing at a scene change works — if you stop at the end of a section, the change now happens there. Before, it waited until you started speaking again, which is exactly when you did not want it
• Markers written mid-sentence are unchanged — those still fire on the word right after them, so you can place a cue precisely inside a sentence

✦ Fixes

• Short lines are tracked again — headings and one-line sentences could never be matched, so the highlight parked at the end of the line before and jumped over them
• Losing your microphone no longer wedges the app — unplug a mic mid-session and you can simply start again, instead of being told listening is already running with no way to stop it
• Stop always stops the scroll — if auto-scroll was started from the OBS dock or your phone, pressing Stop anywhere else did nothing while the script kept moving
• The second display keeps up — minimising the main window used to freeze the detached display on the last word
• Your microphone stays your microphone — VoxiPrompter now remembers it by name, so pairing a headset no longer silently switches you to a different mic
• Speed slider works on the phone, and quickly stopping and restarting no longer leaves every screen showing Idle while tracking runs
• Keyboard shortcuts stay out of dialogs — pressing Space in a confirmation dialog no longer toggles listening behind it
• Clearer port conflicts — if a saved port is busy, it is logged and your saved preference is kept instead of being quietly overwritten

✦ Fixes

• Two markers on one line now both work — a line like [DECK 1] [SCENE: Kamera] used to be read as a single broken marker, so the scene change never happened. Worse, a line like [SCENE: X] your words here [SHOW: Y] dropped those words from tracking, and the highlight stayed off by that many words for the rest of the script
• VoxiPrompter starts without a network — offline, in airplane mode, or on a locked-down venue network, the app used to fail to open at all. It now starts normally; only phone access waits for a network
• A malformed message can no longer close the app — an unusual message from a connected page or integration could shut everything down at once: window, second display, phone, and OBS overlays. It is now logged and ignored

✦ Fixes

• Going back only re-applies where you are — jumping back a paragraph used to fire every marker from the start of your script at once, flicking OBS through every earlier scene before settling. Now only the passage you land on is re-applied, and the further into a long script you are, the more this matters
• Jumping ahead is unchanged — skipping forward still applies each marker you passed, in order, so nothing you set up along the way is missed

✦ What's New

• New [DELAY: 1500] marker — put it before an OBS marker and that change waits the given milliseconds. Switch to a scene, let the transition finish, then bring in your lower third — all from one spot in your script
• It follows you — if you go back to re-read a passage, anything still waiting is cancelled rather than arriving late
• OBS only — web slides driven by cues are never held back, so your deck stays in step with your voice

✦ Fixes

• Honest OBS connection status — closing or restarting OBS is detected within a second and reconnecting starts immediately. Before, VoxiPrompter kept showing "connected" until the next scene marker came along, and that marker was lost
• Lighter connection — VoxiPrompter now tells OBS it does not want event updates it never used, so there is less chatter between the two

✦ Improvements

• No repeated scene switches — if a marker asks for the scene OBS is already showing, VoxiPrompter leaves it alone. Going back over a passage to re-read it no longer sends the same scene change again and again
• Different markers still work as before — only an immediate repeat is skipped, so alternating between scenes or showing and hiding a source behaves exactly as you wrote it

✦ Fixes

• OBS stays connected between sessions — once you have connected to OBS, VoxiPrompter reconnects by itself every time you open the app. No more wondering why your scene markers stopped working after a restart
• A wrong scene name no longer breaks the rest — if a marker points at a scene or source that is not there, VoxiPrompter tells you and moves on, instead of dropping the OBS connection and missing the markers that follow
• Connection settings keep your choice — editing the host or password no longer quietly switches off automatic reconnecting

✦ What's New

• Cues for your slides and tools — write a note in square brackets, like [next slide], and VoxiPrompter signals it the moment you reach that point. Your web slide deck can advance itself while you speak, with no clicker and no assistant. Cues stay hidden from the teleprompter, so nothing changes about what you read
• Works with what you already write — you choose the wording; any bracketed note can be a cue, and your existing OBS scene markers now signal your slides too
• Steadier cue timing — going back to re-read a paragraph no longer leaves your cues stuck; they pick up again from where you are

✦ What's New

• Responsive window layout — the app now adapts smoothly when you resize the window; the sidebar collapses to icons at smaller sizes instead of getting cut off
• Free / Pro edition badge — the header now clearly shows whether you're using the Free or Pro edition
• Faster download page — the download page loads more reliably without depending on external API limits
• Stability improvements — general fixes for smoother day-to-day use