NEW — Name Called on Apple Watch. When someone says your name, your watch taps three times and shows an arrow toward the voice. When the phone can't tell left from right, a fan shows both sides. It works with your iPhone locked while the Vigilant Ear app is open on your watch. Name Called needs Speaker Mode on. NEW — Try it before you need it: a Test button under Preferences → Name Called sends a practice call, and a new tour chapter shows how to set up the watch. BETTER — A second call isn't missed. After an alert, Name Called is ready again in 3 seconds instead of a minute. NEW — Share Vigilant Ear from the help panel (the life ring): a code someone beside you can scan to get the app, a link to the App Store, and our Instagram. BETTER — Sound alerts are less likely to be cut off while Vigilant Ear runs in the background. On iPhones with the enhanced captions, the app now lets go of its caption models after 30 seconds in the background: on an iPhone 17 its memory fell from about 1.6 GB to 85 MB, so iOS is far less likely to close it when another app needs room. Captions come back about a second after you return. BETTER — On iPhones with the enhanced captions, long stretches of speech in a noisy room are captioned more accurately: in a busy room on an iPhone 17, Spanish went from 9.2% of words wrong to 7.3%, French from 25.1% to 22.3%, German from 18.7% to 17.1%. FIXED — Hindi captions no longer slip into Urdu script: on an iPhone 17 in a quiet room, words in the wrong script fell from 26% to 8%, and word errors from 45% to 35%. BETTER — Fewer captions in the wrong language. On iPhones with the enhanced captions, when a speaker is briefly mistaken for another language, the app now checks the words before switching. In a noisy English test recording on an iPhone 17, a 27-second stretch that had been captioned as Romanian stayed in English, in both test runs. This works when the app is captioning English, Spanish, Portuguese, French, German, Italian, Russian or Romanian. FIXED — Captions keep going while your iPhone downloads a translation language. BETTER — Traditional Chinese addresses you with 你, as Apple Taiwan does. FIXED — Translated weather warnings keep their official titles intact. FIXED — In the Acoustic Scope's Spectrogram, sound labels now stay for the whole picture instead of disappearing halfway across. And, miscellaneous improvements to base functionality.

NEW — The map now shows a distance scale at the bottom, always visible, so you can tell how far away things are. NEW — Wind on the map: whenever the weather radar is on, light arrows show which way the wind is blowing, anywhere in the world, updated every hour. BETTER — Captions in loud places. On iPhones with the enhanced captions, an improved speech model makes far fewer mistakes in noise: in a loud bar on an iPhone 17, German captions went from 45% of words wrong to 20%, and Russian from 19% of characters wrong to 9%. Quiet rooms gain too: German from 8.4% to 6.4%, Russian from 5.6% to 3.2%. The same model captions English, Spanish, French, Italian and Portuguese. It replaces the old one by itself over Wi-Fi (about 730 MB). BETTER — The first word of a sentence is no longer cut off, and captions are more accurate overall: on an iPhone 17, word errors fell from 6.7% to 5.4% in English and from 10.5% to 8.9% in German. FIXED — Long captioning sessions no longer use more and more memory (about 135 MB more each hour before). FIXED — A phone call that takes the microphone no longer closes Vigilant Ear. Listening comes back by itself when the call ends. BETTER — Weather warnings show their official title in your language. In Arabic, a "Tropical Storm Watch" is no longer translated word for word into a wristwatch. NEXT in the warning panel steps through several warnings. BETTER — Arabic captions are far more accurate on iPhones with at least 6 GB of memory that don't use the enhanced captions: the same recording went from 43% to 24% word errors on an iPhone 17. Hindi captions spell better there too, with character errors down from 25% to 15%. BETTER — On an iPhone set to another country's region, German, French and Italian captions now use their own country's speech recognizer. German word errors fell from 25% to 9%. FIXED — The last sentence before a pause no longer goes missing. BETTER — Vigilant Ear does far less work when nothing is happening: about 80% less on the main thread at idle on iPhone 16 and 17. BETTER — Traditional Chinese uses Taiwan wording, and several Russian, Arabic and French labels are corrected. BETTER — Remote Link's video is bigger, and captions stay clear of it. IMPORTANT — Linked phones (Constellation) are now always encrypted. Phones running Vigilant Ear 1.1.6 or older need to update to join. And, miscellaneous improvements to base functionality.

BETTER — 30-second Rewind keeps every line again. On iPhones using the captioning model downloaded in 1.1.10, Rewind showed only the line still being spoken. Finished lines and their translations now stay, and a translated line gets a full 30 seconds from the moment its translation arrives. BETTER — Sound direction now uses the exact size of your iPhone on every model that runs iOS 26, from iPhone 11 and SE to iPhone 18 Pro Max. Many models were using an average. BETTER — Weather radar no longer disappears for up to five minutes when a radar service withdraws a frame, and rain no longer looks heavier while a zoom finishes loading. BETTER — Chinese captions no longer put a space between characters, and they end with 。 instead of an English period. Arabic and Hindi captions no longer end with an extra period. And, miscellaneous improvements to base functionality.

NEW — Weather radar on the map. Radar shows rain and snow around you on application startup, from the national radars of the United States, Canada, Japan, Italy and much of Europe, and a satellite rain estimate across most of the rest of the world. If you don't want it for that session, tap the weather pill to turn it off. Touch the screen and the last hour plays back three times; a scale at the bottom reads from Light to Dense. BETTER — More accurate captions, in each speaker's own language. On iPhones with at least 8GB of memory, a one-time captioning model download (about 1 GB, over Wi-Fi, or on cellular after asking) captions each turn in the language actually spoken — so a room that switches between English, Spanish, French and Arabic stays readable. In our tests, word errors in a four-language conversation fell from about 80% to 14%. NEW — Translation while people are still talking. Finished sentences are translated as soon as they end, instead of waiting for the speaker to stop, and translations now arrive in a fraction of the time. BETTER — Captions no longer repeat a long sentence over and over. BETTER — Being on a phone call no longer prevents Vigilant Ear from starting; listening resumes on its own when the call ends. BETTER — On application startup, Standing Watch tells you how long the app wasn't listening for alerts (how long the mic was muted or our app was swiped away). BETTER — Warnings alert you first and translate after, and their titles arrive in your language. BETTER — Speech recognition is ready the moment the app opens, so the first words are captioned sooner. Earthquake checks keep running every 5 minutes in the background. NEW — A life ring on the map lists who to call for help in the country you are in. NEW — Spoken alerts have their own switch. And, miscellaneous improvements to base functionality.

NEW — Traditional Chinese. The app and its App Store page are now available in Traditional Chinese for readers in Taiwan and Hong Kong, written in Taiwan's own everyday vocabulary rather than converted character by character from Simplified Chinese. NEW — The Acoustic Scope's Partials panel reads music. Partials now shows the harmonic series and each note's name in their own columns, and higher sounds such as smoke alarms are named at their true pitch. You can also set target notes which draws as a line to sing or tune against. NEW — The Acoustic Scope's decibel readings on iPhone are now calibrated against a sound level meter — accurate to about ±2 dB (not a certified instrument). BETTER — The Acoustic Scope runs cool. The Spectrogram now costs next to nothing to leave open, so it will not warm up your phone however long you watch it. BETTER — A calmer Acoustic Scope. The Spectrogram and Partials views redraw at a steady, readable pace instead of racing — without missing any of what they hear. FIXED — Music ID song titles update more gracefully. FIXED — The "can't be translated" alert for languages Apple's translation engine doesn't cover, such as Romanian, sits at the top of the screen and is minimized but present if you need to see the details. And, miscellaneous improvements to base functionality.

NEW — Romanian. The app now recognises Romanian speech and captions it, and the interface itself is fully translated into Romanian. One limit worth knowing up front: Apple does not offer Romanian in its translation engine, so Romanian captions appear but cannot be translated into any other language — nor can other languages be translated into Romanian. BETTER — The language detector works better because we feed it better raw speech data. All on-device. BETTER — Bilingual conversations are improved. A single voice using a primary language but adding words from another mid-sentence used to confuse our detector. It now averages the context so the language stays locked even if non-primary language words are used. BETTER — Translation quality. Captions and weather alerts were using two different translation engines, one tuned for speed. Everything now uses the same higher-quality engine. And, when a translation engine's AI model has to download, you see that it is downloading rather than watching untranslated text and wondering whether the feature is broken. FIXED — "Name Called" feature now points correctly at whoever said your name. With more than one person nearby, the arrow could point at whoever the app happened to be captioning already rather than at the person who spoke — placing the one event this feature exists to place on the wrong person. The direction is now recorded at the moment they called your name. As before, no direction is shown if we are unsure. FIXED — Weather alert notifications say what happened. The line under the title was the same generic sentence on every alert. It now carries the event itself — what was issued and until when. If you only ever see the banner, that line is the whole alert. BETTER — Distances follow your region, not your language. Miles or kilometres now track where you are rather than the language you read the app in, so a Spanish-reading user in the United States gets miles. This applies in the Acoustic Scope, the tutorial range rings, and throughout. BETTER — Right-to-left languages. Several screens have been corrected for Arabic readers. And, miscellaneous improvements to base functionality.

NEW — "Name Called". The app can tell you when someone like a barista says your name, and which direction it came from. Add the names you want watched: your own, a child's, or a family member. When one turns up in the live captions, your phone and watch vibrate to alert you and show the direction the voice came from. "Someone was talking TO you" is a different event from "someone spoke", and it is the one you cannot recover afterwards. We only show the direction if we accurately measured it. NEW — "Mask Strong Language". Inappropriate words arrive in captions as plain text with nothing to soften the blow. With a new toggle turned on, the app replaces those words with symbols. This is permanently on for anyone set up as a minor but it is also a switch any other user can turn on. The filtering is determined by iOS rather than our word list. So, it works in your chosen language. NEW — "Word Emphasis". The word a speaker leaned on is now marked in the caption so that Deaf/HoH users can finally see that. Spoken aloud, "I didn't say THAT" and "I didn't say that" are two different sentences. An identical transcript lost the subtle meaning completely. The app listens for a word that the speaker's voice lifted with a change in pitch and shows it in bold. At most one word per line and nothing at all when the app is unsure. Certain languages that naturally use pitch changes are excluded, such as Chinese and Japanese. NEW — Turkish, our thirteenth language. The whole interface is translated. BETTER — Captions are substantially more accurate. The app used to run a separate "person" recognizer for each voice, so whenever it changed its mind about who was speaking — which could happen in the middle of sentences — the transcription restarted, and words were lost or landed in the wrong line. One recognizer now runs continuously and the speaker identifications are attached later. BETTER — Privacy. The "Constellation" link between your own devices now uses a unique, one-time secret key that exists only for that connection and is destroyed when it closes. BETTER — You can end a "Remote Link" from the main display. Hanging up used to mean an unintuitive trip through a settings panel. FIXED — The app now reads in the language you picked, everywhere. If the phone's language and the app's language differed, some sections quietly fell back to the phone's — such as the explanations under some preference panel toggles, "Why Vigilant Ear?", and the Acoustic Scope tutorial. FIXED — The "Standing Watch" bubble used to say "beside you" in English only. It is now translated into all of our thirteen supported languages.

FIXED — A ticking clock is no longer mistaken for music. A quiet room with a steady repeating sound in it — a wall clock, a fan, a hum — could tip the app into thinking music was playing, which changed how it treated everything else it heard. The app now checks a second opinion before it believes music is present, and that opinion covers the whole family of musical sounds rather than just a handful. Real music is still picked up, and still identified. BETTER — Who is speaking is separated more accurately. Until now the app decided a new person had started talking by watching for a pause. Pauses are a poor signal: people talk over each other, and a thoughtful speaker pauses mid-sentence. A second AI model now listens specifically for the moment the voice changes, and the old pause-based method stays underneath it as a fallback if that model has nothing to say. The result is cleaner separation of one person's words from the next person's. BETTER — The bar for "this is speech, not silence" is now read from the room. It used to be a fixed number chosen in advance, which was too high in a quiet room and too low in a noisy one. The app now sets it from the audio actually in front of it. BETTER — Naming a voice refuses a recording that is not worth keeping. If the sample it captured is too short or too muddled to recognise you by later, the app says so at the time instead of saving something that will not work. FIXED — Fewer false "check your device" warnings. That alert now requires both of the app's sound classifiers to have gone quiet, not just one of them. Plus: the speaker-separation model ships inside the app rather than downloading itself later, so there is nothing to fetch on first run and it works with no network at all. That is most of why this update is larger than usual.

FIXED — Captions no longer stop when the speaker changes. On the newest iOS, roughly half of all handoffs from one talker to the next left captions frozen mid-conversation. The app was running out of speech-recognition slots and never getting them back. It now waits for the old slot to be returned before opening the next one, and recovers if it isn't. FIXED — Remote Link shows captions on the other phone. The captions this phone produces were meant to travel to the linked phone and were not arriving. They arrive now. FIXED — A Remote Link call no longer ends itself after a few minutes. The app was treating the expiry of the short-lived invitation code as the call ending. A connected call ignores it. BETTER — Song identification remembers the room. The app learns how quiet a room is before it will believe music is playing. It now remembers that for the handful of places you listen most, so it doesn't have to relearn the room every time you open the app. BETTER — Background sound detection re-arms itself. If the system stops handing us sound while the app is in the background, the app widens its listening window to get itself back — and tells you if that doesn't work, instead of sitting there quietly. BETTER — Sharper direction and distance on the newest iPhones. iPhone 17 Pro, iPhone 17 Pro Max and iPhone 18 Pro now use their own microphone spacing instead of a generic figure. Phones we haven't measured yet get a closer default too. Plus: the Remote Link pause controls and the "Sound alert" notification placeholder now speak all twelve languages. And a speaker's colour dot holds gray for a moment while the app is deciding who is talking, rather than flashing the wrong person's colour and correcting itself.

BETTER: Remote Link shows captions from the linked phone to yours, and translates if necessary to your chosen language. And, you can pause the link on your end to temporarily stop video and captions at any time, and then resume. BETTER: Display fixes, notifications, and (automatic) wider connection options for Remote Link.

NEW — Hindi. Vigilant Ear is now fully translated into Hindi, our twelfth language. NEW — India severe-weather warnings, from NDMA Sachet, in Hindi and English. NEW — Remote Link. Reach someone who isn't with you, using video and text — the thing a phone call would normally do. No audio is used at any point, so nothing depends on hearing at either end, and it gives you a way to sign with someone. You send an invitation code; they join without needing Power Pack+. FIXED — Weather and emergency alerts now reach you in the language you read. Alerts from Japan, Korea, China and Brazil were arriving exactly as published — in Japanese, Korean, Chinese and Portuguese. Now fixed. BETTER — Who said what. In a three-person conversation, shorter turns were being folded into the previous speaker's voice. Each turn now keeps its own speaker. BETTER — The confidence marks on captions are calibrated from measurement instead of an estimate. A marked word is now much more likely to be one that is genuinely wrong, and the unmarked words are safer to trust. BETTER — Song identification listens for the music actually changing instead of checking on a timer. Titles keep up with the track — including continuous DJ mixes with no gaps between songs — and a title no longer blinks out and back while the song is still playing. FIXED — The identify-this-song button responds the moment you press it. FIXED — Acoustic Scope shows real sound again after a trial ends, instead of "listening…" forever. Plus: the song title and the sound HUD now move out of the way when the keyboard is open, for a Constellation message send, in both portrait and landscape.

NEW — Your Constellation-linked phones now live on the location independent ring in the middle of the map. They used to be location-pinned so if you moved out of range, you lost the ability to interact with them. BETTER — Captions that read like a conversation. When one person keeps talking, their words now grow in a single row instead of starting a new bubble every few seconds. Much less flicker, much easier to follow. BETTER — We tidy up speaker caption lines using on-device machine learning in all languages. BETTER — Multiple iPhones, more of the sentence. When meshed phones hear the same speech, they merge the best choices to reveal better transcription. And, thus, translations or spoken captions. FIXED — Less resource heavy during background sound detection, which is kinder to the battery. Plus: every practice alert in the Feature Playground was updated. And linked iPhones keep a tighter shared clock to improve merge caption quality.

Captions • Better accuracy in difficult rooms. Captions now favour getting the words right over getting them fast. • Corrections now work in Japanese, Chinese, Korean and Arabic. Vigilant Ear reviews a caption line after it first appears and improves it. For these languages that step had never once applied — the check that approves a correction was measuring text in a way that does not fit these writing systems, so every correction was discarded. It now works. • Names and song titles. People you are linked with, and the song playing in the room, are recognised more reliably. Constellation (two or more phones) • Captions now combine what each phone heard. The phone nearest a speaker usually hears best, and words that only one phone caught are no longer lost. Weather • Japan: earlier notice. Vigilant Ear now reads the Japan Meteorological Agency's advance bulletins — the plain-language notices issued days before a typhoon or heavy rain — not only the warnings issued once the danger has already arrived. It also reads the sudden-downpour and landslide bulletins. • Advance notices arrive quietly. They appear on the map and in your notifications without the sound and vibration reserved for a real warning.

PRIVACY — Weather warnings no longer come straight from foreign governments' servers. Vigilant Ear now fetches every official alert through its own service, so your phone never contacts the American, Japanese, Korean, Chinese, Australian, Canadian, Brazilian or European weather agencies directly — none of them sees your device or how often you check. The exact "am I inside this warning?" test still happens on your phone alone and never leaves it. FIXED — Warnings that had already been lifted could stay on screen. In Korea, nearly every warning shown could be one that had already been cancelled. Cancellations are now read and acted on for every country. FIXED — Flood warnings could be missed in Australia. One bulletin can cover nine separate river areas, and only the first was being read. FIXED — Alerts in China could appear in the wrong place — some as far as 170 km from the county they were issued for. FIXED — Repeat notifications. In Australia a single warning could notify you 64 times as it was re-issued. Japan had the opposite problem and stayed silent when a warning escalated to the highest level. SIMPLER — One switch for weather. The nine per-country switches are gone; the app already watches whichever country you're standing in. FIXED — Earthquakes have their own switch. Turning weather off used to silence earthquake alerts too, even while the Earthquakes switch still read "on".

BETTER — Captions quietly fix themselves. Right after a sentence lands, the app re-checks it against a cleaner read of the audio and can restore missed or misheard words within a couple of seconds. FIXED — A real siren passing by while music played could be missed. Sirens now break through the music filter. BETTER — One voice stays one voice. Playing music used to split a single speaker into two or three phantom "voices." We rebuilt how voices are identified around music — in our field tests: zero phantom voices, every speaker kept their own color and captions. And when two people talk over each other, the app now separates the voices on-device to keep each on its own caption track. BETTER — Music can't fool the app. It now remembers what your room sounds like when it's quiet, so a whole evening of music can no longer wear that memory down. In our overnight test it caught every real song change — and wastes far fewer lookups on songs it already named. BETTER — Captions hold steady in noisy places. The app now confirms two people are really talking before it separates voices — and in truly loud rooms it never degrades your captions trying. BETTER — Who's-talking labels you can trust: a voice is only confirmed as a real person from independent evidence. BETTER — Weather alert feed accuracy and notifications are improved, along with alert scrolling when more than one alert is in your area. European alerts now arrive through our new alert service — faster and more reliable. BETTER — Some in-app text is more accurate for your chosen language. Plus safer custom sound-pack imports and many small fixes.

NEW — Standing Watch. A small lamp now stands at the top of your screen, always on watch: cyan means a steady room, amber means something changed — a new voice, the room going quiet, or something approaching, with a direction when we can measure one. Red you already know. It counts voices that are speaking, never silent people. NEW — Parla italiano! The whole app now speaks Italian — tours and live captions included. NEW — Brazil severe-weather alerts (INMET), in Portuguese for Portuguese readers. BETTER — Who's talking. Voice identification is faster to notice a second voice and harder to fool in noisy rooms, and named voices are never crowded out. Every voice's dot now carries its number, so two similar colors can never blur two people together. BETTER — Battery & memory. The app travels much lighter in the background and returns instantly. Plus portrait polish everywhere: a full-width threat strip, smarter dialogs, tidier tutorials.

NEW — Portrait. Since day one, Vigilant Ear has made you turn your phone sideways. No more. Hold it the way you actually hold a phone: the map runs tall, the title sits under the notch, panels open full width, and the sound picker blooms straight down. Landscape is still there — it's the dense dashboard it always was. Now it's your choice. NEW — The first sentence gets caught. The hardest words for any captioning app are the first ones after a quiet room — and for a Deaf user, "the smoke alarm is going off" is usually a first sentence. We found three separate places where those opening words could be delayed, clipped, or lost entirely, and fixed all three. Speech that arrives out of silence now gets transcribed whole, even when the app has to catch up to it. IMPROVED — Rewind reads top-down now. The newest line pops in at the top and older lines age off the bottom, so you never hunt for "just now." FIXED — Tutorial progress finally sticks. Chapter rings fill as you go, stay filled across app updates, and only reset when a chapter's content actually changes — because if we rewrote the tour, it's worth a fresh look. Plus smaller portrait touches everywhere: captions that never jump when an emergency pill appears, a Shazam gauge you can actually see, and a compass that stopped hiding behind your battery indicator.

NEW — Captions that tell you how sure they are. Every line now carries a small three-bar mark: three bars means the app heard it clearly, one bar means take it with a grain of salt. Individual words it wasn't confident about are underlined. No other captioning app admits its own uncertainty — and in a noisy room, knowing which words to trust matters as much as the words themselves. NEW — Rewind. Missed what someone just said? One tap brings back the last thirty seconds of captions — the lines that already scrolled off your screen. Text only: no sound is kept, and nothing is ever saved to your phone. NEW — Conversations. Show a QR code, a friend scans it, and their phone's captions join yours — their words on your screen, with their name on them. Private, over your local network, no cloud. NEW — Direction Tones. If you hear from only one ear, sounds arrive without a direction. With earphones in, a short tone in your hearing ear says where a sound came from: three beeps is ahead, two is on your non-hearing side, one is behind you. Headphones only, and off until you turn it on. Better captions in Japanese, Korean, Russian and Arabic, and Caption Bench now shows what's typical for each language, so a normal score doesn't look like a failure. Plus: a heads-up if iOS sound recognition stops behaving, so a quiet phone never looks like a working one.

NEW — The Acoustic Visualizer: music you can see. Every beat detonates a ring of light, the bass breathes, the highs rain sparkles, and the song title rides a rolling globe. Set your DJ name and colors (rainbow, pride, and more) in Preferences, then put the whole show on a TV with the scope's new Mirror button — cable or AirPlay. NEW — Caption Bench: test caption accuracy in your language, right on your phone (Playground → Quality Check). Captions are now far more reliable: words no longer vanish when the app switches listening sessions mid-sentence, and speaker identity holds steadier through pauses. The Acoustic Scope now opens full screen.

Japan and Canada now have official severe-weather alerts, from their own national agencies. JAPAN — JMA (気象庁) Warnings and Emergency Warnings from the Japan Meteorological Agency, for the forecast region you are actually in: storm, high waves, heavy rain, flood, heavy snow and the rest, in the agency's own wording. Japanese readers get JMA's Japanese; everyone else gets JMA's official English. Advisories (注意報) are left out on purpose — they stand over much of the country most days, and Vigilant Ear should only interrupt you for the levels that mean act now. CANADA — ECCC Warnings and watches from Environment and Climate Change Canada, in English or French, matched to the alert zone you are standing in rather than a box drawn around the province. ALSO IN THIS RELEASE • Weather alerts are matched to the real shape of your region, so a warning meant for the next prefecture over no longer reaches you. • Alert notifications now lead with the alert itself rather than the issuing agency's acronym, in every language. Severe-weather alerts are free, and stay free. Each country's feed only runs where it applies, so Japan's warnings stay quiet in Toronto and Canada's stay quiet in Osaka. As always: listening, captions, and translation run entirely on your device.

Meet Witness Ear — an optional 24-hour sound journal — plus live translations that now happen practically instantaneously. WITNESS EAR (NEW) Keep a 24-hour journal of the sounds around you. While it's on, classifications — with time, sound level, and where your phone was — stay on your phone for up to 24 hours, ready to export as a clean PDF report: sounds grouped into families, activity by hour, episodes, locations, and an integrity fingerprint you can quote. With Constellation, linked phones contribute what they heard too — one report from the whole array, each observation labeled by phone. Off by default, nothing is ever uploaded, and the report itself speaks all nine app languages. Find it in Preferences → Sound Journal; a full guide lives in the menu. TRANSLATIONS, NOW INSTANT Live caption translation moved to the fastest on-device engine. Sentences that used to take a few seconds now translate in a fraction of one — the French (or any language) appears almost as the words settle. Spoken translations through headphones keep pace with real conversation: in order, uncut, and never in the wrong language — when a translation isn't ready, Vigilant Ear stays quiet rather than guessing out loud. WORLD EVENTS • Earthquakes: official USGS confirmations minutes after the shaking — what it was, how strong, how far. A confirmation, not an early warning. • Deep Rumbles — Research Array: a Map button opens the live world map of low-frequency events heard by the research array. Viewing shares nothing. ALSO IN THIS RELEASE • Acoustic Scope no longer freezes after the app returns from the background. • A music-tone indicator on the Partials view. • Translations completed across every corner of the app. As always: listening, captions, and translation run entirely on your device.

Russian — the tenth language. • Live captions, translation, and the entire interface now work in Russian. • Spoken-language detection is more accurate, and switches faster when someone starts speaking another language. • Captions now arrive smoothly instead of in bursts. • Fixed: the language readiness indicators in Preferences reported the wrong state for some languages. • Fixed: the app no longer downloads a language you didn't ask for. • Feature Playground: Russian added to the auto-detect and live-translation demos.

Vigilant Ear is now a native iPad app. • On iPad and other large screens, live captions move into a see-through panel beside the map — more conversation history, captions that stay up twice as long, and the map stays visible and centered underneath. The panel tucks away when nobody is talking. • iPad keeps listening in every orientation. • Fixed: live captions could fail to start on devices that had reached the on-device speech-model limit. • Fixed: automatic caption translation could be silently disabled on some installs — it now works out of the box. • Clearer demos and small polish throughout.

• Spoken captions: Vigilant Ear can now speak live transcriptions aloud through your Bluetooth headphones or hearing devices — free for everyone. • Deep rumbles: a new low-frequency awareness sense helps confirm officially reported earthquakes faster. Always a confirmation, never a prediction. • Smarter alerts: detections now cross-check multiple evidence sources — fewer false alarms, and sirens are recognized reliably even with music playing. • Street-aware tracking: passing vehicles lock onto the road they're actually on, with cleaner distance readouts. • Sound AR: unified markers, off-screen direction indicators, and smart clustering when things get busy. • New dedicated car-alarm alert. • Constellation mesh reliability improvements, tutorial updates, and many fixes.

Vigilant Ear keeps getting better. • Speaker captions now pop in faster to reduce missed words, and stay live through pauses in conversation. • Enhanced recognition of emergency alerts such as babies crying, sirens, and alarms. • New earthquake alerts — feel a buzz and see the area that felt it on your map when a quake is reported nearby. • Faster, more accurate on-device (no cloud) language translation, with automatic language detection. • Display enhancements and polish throughout.