chore(deps): bump cryptography from 49.0.0 to 50.0.0 in the pip group across 1 directory

.github/CODE_OF_CONDUCT.md

@@ -46,7 +46,7 @@ an individual is officially representing the community in public spaces.
 
 Instances of abusive, harassing, or otherwise unacceptable behavior may be
 reported to the project maintainer through **<https://voxislive.com>** or via a
-private [GitHub security advisory](https://github.com/DavutAkca/voxislive/security/advisories/new).
+private [GitHub security advisory](https://github.com/VoxisLive/voxislive/security/advisories/new).
 All complaints will be reviewed and investigated promptly and fairly. Maintainers
 are obligated to respect the privacy and security of the reporter of any incident.
 

.github/CONTRIBUTING.md

@@ -11,7 +11,7 @@ and may be incorporated with attribution in the project history.
 ## Development setup (BYOK / developer build)
 
 ```powershell
-git clone https://github.com/DavutAkca/voxislive.git
+git clone https://github.com/VoxisLive/voxislive.git
 cd voxislive
 python -m venv .venv            # Python 3.11-3.13, 64-bit (3.14 not supported yet)
 .\.venv\Scripts\Activate.ps1

.github/ISSUE_TEMPLATE/config.yml

@@ -4,5 +4,5 @@ contact_links:
     url: https://voxislive.com/licensing
     about: Commercial use, white-label, or the official SaaS build.
   - name: Report a security vulnerability
-    url: https://github.com/DavutAkca/voxislive/security/advisories/new
+    url: https://github.com/VoxisLive/voxislive/security/advisories/new
     about: Please report security issues privately, not as a public issue.

.github/SECURITY.md

@@ -5,7 +5,7 @@
 Please **do not** open a public issue for security problems.
 
 Report privately through GitHub's
-[**Report a vulnerability**](https://github.com/DavutAkca/voxislive/security/advisories/new)
+[**Report a vulnerability**](https://github.com/VoxisLive/voxislive/security/advisories/new)
 form (the repository's **Security → Advisories** tab). We aim to acknowledge a
 valid report within a few days and will coordinate a fix and disclosure with you.
 

.github/dependabot.yml

@@ -0,0 +1,11 @@
+version: 2
+updates:
+  - package-ecosystem: "pip"
+    directory: "/"
+    schedule:
+      interval: "weekly"
+
+  - package-ecosystem: "github-actions"
+    directory: "/"
+    schedule:
+      interval: "weekly"

.github/workflows/quality.yml

@@ -21,15 +21,15 @@ jobs:
       matrix:
         python-version: ["3.11", "3.13"]
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v7
 
-      - uses: actions/setup-python@v5
+      - uses: actions/setup-python@v7
         with:
           python-version: ${{ matrix.python-version }}
           cache: pip
           cache-dependency-path: requirements.lock
 
-      - uses: actions/setup-node@v4
+      - uses: actions/setup-node@v7
         with:
           node-version: "22"
 
@@ -56,9 +56,9 @@ jobs:
     needs: test
     runs-on: windows-latest
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v7
 
-      - uses: actions/setup-python@v5
+      - uses: actions/setup-python@v7
         with:
           python-version: "3.13"
           cache: pip

.github/workflows/release-hygiene.yml

@@ -16,10 +16,10 @@ jobs:
   hygiene:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v7
         with:
           fetch-depth: 0  # full history so the --history scan is meaningful
-      - uses: actions/setup-python@v5
+      - uses: actions/setup-python@v7
         with:
           python-version: "3.12"
       - name: Scan tracked tree and history for leaks

CHANGELOG.md

@@ -6,6 +6,99 @@ Notable changes to Voxis. Version bumps are tagged in commit messages
 ## Unreleased
 
 ### Fixed
+- A Qwen capacity error ("thread pool exhausted") that used to be retried up
+  to 8 times with fast backoff is now recognized as terminal instead of
+  generic-transient — it was hammering an already-saturated shared rate limit
+  and prolonging the very outage it was reacting to. A terminal error now
+  gets one immediate retry on a server-provided sibling connection pool
+  before falling back to the alternate engine.
+- On VB-CABLE setups, the volume mirror assumed the physical playback device
+  sat at 0 dB. If it didn't — the common case — that device's own level
+  silently capped the output under whatever the mirror computed: the display
+  showed full volume with nothing left to raise while the actual level was
+  quieter. Voxis now raises the physical device to full for the session and
+  restores it on stop.
+
+### Added
+- The top bar now shows which version you're running, and flags it when a
+  newer one is available — checked in the background against a small public
+  manifest (`voxislive.com/app.json`) that also carries language coverage for
+  the website. The check is unauthenticated and sends no user data, so it
+  runs on the OSS/BYOK build too, not just the official release.
+- Voxis says something when it cannot hear anything. If no sound reaches it in
+  the first seconds of a Video session, it now tells you — and names the two
+  things that actually cause it: nothing is playing, or you are speaking into
+  your microphone, which Video mode does not translate. Until now the app stayed
+  completely silent in that state; measurement across every recorded session
+  showed 40% of them ended inside 30 seconds, with another attempt following a
+  median 14 seconds later, i.e. people were retrying, not losing interest.
+- The gap between pressing Start and the first translated word is no longer a
+  blank screen: the stream shows that it is connecting, escalates if the
+  handshake drags past ten seconds, and sets the expectation that the first
+  translation lands a few seconds after someone speaks. The line explaining that
+  delay used to appear only with the first caption — after the wait it was meant
+  to explain, and never at all for anyone who gave up first.
+- Release notes now cover every version you skipped, not just the running one.
+  Store updates land in the background and can jump several versions at once,
+  which is how 1.0.50's notes reached nobody who went from 1.0.49 to 1.0.51 in a
+  single update. The card also links to the full changelog on the website.
+- Translated-voice gender, chosen per direction: "Translation voice" (what you
+  hear) and "My voice" (what the other person hears you as, in Meeting mode).
+  Both live in Settings › Translation — it is set once for a meeting, not
+  adjusted per session, so it does not belong on the main screen. Available
+  on the 31 targets served by the Qwen engine — measurement showed the Gemini
+  translate model ignores voice selection entirely, so on the remaining targets
+  the picker says so instead of silently doing nothing. Until now every session
+  used one server-default voice, which is female.
+- Release notes in the app: after an update you see what changed once, in your
+  own UI language, instead of having to find the Microsoft Store listing. A
+  fresh install is not shown a changelog — it gets the onboarding tour.
+- A ready-made list of ~31 proper nouns ships with the term box, so brand and
+  product names are spelled right before anyone types anything. It can be
+  switched off, and your own terms always win over a shipped one.
+
+### Changed
+- The term box moved out of General into its own Settings › Translation tab and
+  is no longer called "Meeting terms": it has always applied to Video mode too.
+
+### Fixed
+- A translator session that connected but never produced any output no longer
+  looped retrying the same dead connection forever: the retry counter was
+  reset by time-alive alone, so a connection that stayed open past 5 seconds
+  before dying with zero output kept "proving the path works" and never
+  reached the threshold that hands the session to a backup engine. It now
+  also requires that the session actually produced output before resetting.
+  Paid plans additionally fail over to the backup engine on the very first
+  dropped connection instead of riding out several retries.
+- A session no longer goes permanently silent when the translation service
+  starts rejecting the audio it is sent. Those rejections still count as
+  "the engine is hearing us", so both self-heal watchdogs stayed disarmed: the
+  session kept its connection, showed as active, produced nothing, and never
+  reconnected. One field meeting lost its last 13 minutes that way. Voxis now
+  treats a rejected utterance as input and reconnects itself.
+- The close button of the translation-history window sits at the top right,
+  like every other window, instead of next to the title.
+- Meeting mode no longer translates its own outgoing voice back at you. On a PC
+  with VB-Cable installed, both directions shared the one cable, so the
+  translation Voxis had just sent into the call came back in as if the other
+  party had said it — a third, phantom voice in your own language.
+- Captions no longer run words together at a sentence boundary
+  ("bekleyelim.Bu arada," instead of "bekleyelim. Bu arada,") — it affected
+  a quarter of all caption lines on the Qwen engine.
+- Repeated caption text is removed: both a whole line re-spoken as the next
+  turn, and a clause the engine emits twice inside one line. The translated
+  speech only ever said it once.
+- A caption line no longer swallows 20+ seconds of speech; past a length or
+  time budget it splits at the next sentence end.
+- Subtitle exports (SRT/VTT) are wrapped to a readable line width — cues used
+  to run off the frame, the longest being a single 548-character line.
+- Subtitle timings now start from the session, not from the first translated
+  word, so an export lines up with its own audio recording. A session that
+  began before anyone spoke shifted every cue.
+- Meeting mode keeps its two directions apart. Both translators fed one caption
+  line and one transcript, so the other party's words and your own interleaved
+  with no way to tell them apart. Each side is now labelled, on screen and in
+  the saved transcript.
 - The "Original audio (while speaking)" control on VB-CABLE setups now ducks
   the original speaker, not the background music — a same-day regression had
   it scaling the whole ambient mix uniformly instead.
@@ -34,6 +127,9 @@ Notable changes to Voxis. Version bumps are tagged in commit messages
   could interrupt a Meeting session.
 
 ### Added
+- Settings > **General**: a **Meeting terms** box. Names your meetings use —
+  companies, products, people — can be listed one per line so the translator
+  spells them correctly instead of guessing.
 - Settings > **Saving** tab: transcript folder, audio recording, and
   automatic TXT/SRT/VTT export formats (generated alongside the JSON on
   every save) are now grouped in one place; quick TXT/SRT/VTT export
@@ -64,7 +160,7 @@ Notable changes to Voxis. Version bumps are tagged in commit messages
 - A language-swap control to exchange the two translation targets in one step.
 
 *Thanks to [Vladimir Vorobyov (@uladzemer)](https://github.com/uladzemer) for
-this contribution — [audio diagnostics fix](https://github.com/DavutAkca/voxislive/commit/9ee5e50c637dd8435402248777f8b959ea0c81cb).*
+this contribution — [audio diagnostics fix](https://github.com/VoxisLive/voxislive/commit/9ee5e50c637dd8435402248777f8b959ea0c81cb).*
 
 ### Fixed
 - A crash partway through starting a Video/Game or Meeting session could leak
@@ -92,4 +188,4 @@ this contribution — [audio diagnostics fix](https://github.com/DavutAkca/voxis
   push and pull request against `main` (Python 3.11 and 3.13, Windows).
 
 *Thanks to [Vladimir Vorobyov (@uladzemer)](https://github.com/uladzemer) for
-this contribution too — [runtime/storage hardening](https://github.com/DavutAkca/voxislive/commit/8cdc51eee8e68726d4c21410fa17332a1ccbf892).*
+this contribution too — [runtime/storage hardening](https://github.com/VoxisLive/voxislive/commit/8cdc51eee8e68726d4c21410fa17332a1ccbf892).*

README.de.md

@@ -5,7 +5,7 @@
 ![Platform](https://img.shields.io/badge/platform-Windows%2010%20%7C%2011-0078D6?logo=windows&logoColor=white)
 ![Python](https://img.shields.io/badge/python-3.11--3.13-3776AB?logo=python&logoColor=white)
 ![License](https://img.shields.io/badge/license-PolyForm%20Noncommercial%201.0.0-blue)
-[![GitHub stars](https://img.shields.io/github/stars/DavutAkca/voxislive?style=social)](https://github.com/DavutAkca/voxislive/stargazers)
+[![GitHub stars](https://img.shields.io/github/stars/VoxisLive/voxislive?style=social)](https://github.com/VoxisLive/voxislive/stargazers)
 
 > Echtzeit-Sprachübersetzung für Windows — übersetze jedes Video, Spiel oder Meeting und höre es live in deiner eigenen Sprache.
 >
@@ -19,7 +19,7 @@
 **📖 Anleitung:** [Entwickler / BYOK-Setup](docs/INSTALL_BYOK.md) — die Endnutzer-App wird über den **Microsoft Store** ausgeliefert; Setup-Doku unter [voxislive.com](https://voxislive.com).
 
 > [!WARNING]
-> **Laden Sie nur von [voxislive.com](https://voxislive.com), dem Microsoft-Store-Eintrag oder diesem Repository (`github.com/DavutAkca/voxislive`) herunter.** Kopien dieses Repositorys wurden auf anderen GitHub-Konten gefunden, einige leiten zu Installern auf fremden Seiten weiter — diese sind **nicht offiziell** und können schädlich sein. Dieses Projekt verteilt niemals einen Installer über eine Drittanbieter-Seite oder ein anderes GitHub-Konto. Wenn Sie eine verdächtige Kopie finden, melden Sie sie bitte: [support@voxislive.com](mailto:support@voxislive.com).
+> **Laden Sie nur von [voxislive.com](https://voxislive.com), dem Microsoft-Store-Eintrag oder diesem Repository (`github.com/VoxisLive/voxislive`) herunter.** Kopien dieses Repositorys wurden auf anderen GitHub-Konten gefunden, einige leiten zu Installern auf fremden Seiten weiter — diese sind **nicht offiziell** und können schädlich sein. Dieses Projekt verteilt niemals einen Installer über eine Drittanbieter-Seite oder ein anderes GitHub-Konto. Wenn Sie eine verdächtige Kopie finden, melden Sie sie bitte: [support@voxislive.com](mailto:support@voxislive.com).
 
 ---
 
@@ -60,7 +60,7 @@ Windows audio ──► Capture ──► Silero VAD gate ──► Gemini Live
 ## Schnellstart (Entwickler-Build)
 
 ```powershell
-git clone https://github.com/DavutAkca/voxislive.git
+git clone https://github.com/VoxisLive/voxislive.git
 cd voxislive
 python -m venv .venv
 .\.venv\Scripts\Activate.ps1
@@ -114,7 +114,7 @@ Die beiden Richtungen haben unterschiedliche Anforderungen:
 | **Eingehend** (du hörst sie in deiner Sprache) | Hört das Systemaudio ab, übersetzt es und gibt es auf deinen Kopfhörern wieder | **Keine zusätzliche Installation** |
 | **Ausgehend** (deine Stimme geht übersetzt hinaus) | Übersetzt dein Mikrofon und speist ein virtuelles Mikrofon | **Ein virtuelles Mikrofon (VB-CABLE) ist erforderlich** |
 
-> Unter Windows ist die einzige Möglichkeit, ein „Mikrofon" bereitzustellen, das eine Meeting-App (Teams/Zoom/Meet) auswählen kann, ein virtueller Audiotreiber — deshalb benötigt die ausgehende Richtung VB-CABLE. Ohne einen solchen laufen Meetings automatisch im **Nur-Zuhören**-Modus (du verstehst sie; deine Stimme geht unübersetzt hinaus).
+> Unter Windows ist die einzige Möglichkeit, ein „Mikrofon" bereitzustellen, das eine Meeting-App (Teams/Zoom/Meet) auswählen kann, ein virtueller Audiotreiber — deshalb benötigt die ausgehende Richtung VB-CABLE. Ohne einen solchen Treiber laufen Meetings automatisch im **Nur-Zuhören**-Modus (Sie hören die Gegenseite übersetzt; Ihre eigene Stimme wird unübersetzt übertragen).
 
 ### 1. VB-CABLE installieren (einmalig, kostenlos)
 1. Lade es von <https://vb-audio.com/Cable/> herunter.
@@ -124,7 +124,7 @@ Die beiden Richtungen haben unterschiedliche Anforderungen:
 ### 2. Voxis konfigurieren
 - Stelle die Sprachen im Panel ein: **Ich höre: Türkisch**, **Für andere: Englisch**.
 - Einstellungen → **Ausgabegerät**: deine echten Kopfhörer · **Mikrofon**: dein echtes Mikrofon — das, in das du sprichst; Voxis hört hier zu.
-- **Das virtuelle Kabel wird automatisch erkannt.** Beim Start findet Voxis ein installiertes Kabel (VB-CABLE / VB-Audio / VoiceMeeter) und richtet das Meeting-Routing selbst ein — kein Bearbeiten von `config.json`.
+- **Das virtuelle Kabel wird automatisch erkannt.** Beim Start erkennt Voxis ein installiertes Kabel (VB-CABLE / VB-Audio / VoiceMeeter) automatisch und konfiguriert das Routing selbstständig — eine manuelle Anpassung der `config.json` ist nicht erforderlich.
 
 ### 3. Die Meeting-App konfigurieren (Teams / Zoom / Meet)
 - Stelle das **Mikrofon** auf **„CABLE Output (VB-Audio Virtual Cable)"** ein — die *Aufnahme*-Seite des Kabels (`CABLE Output`, **nicht** `CABLE Input`). Das ist das Mikrofon der Meeting-App, nicht das in Voxis gewählte echte Mikrofon: Voxis schreibt dein übersetztes Englisch in das Kabel, und die Meeting-App liest es hier zurück.
@@ -215,7 +215,7 @@ Beiträge sind willkommen — indem du einen Pull Request öffnest, erklärst du
 
 ## Support
 
-- **Issues:** [GitHub Issues](https://github.com/DavutAkca/voxislive/issues)
+- **Issues:** [GitHub Issues](https://github.com/VoxisLive/voxislive/issues)
 - **Kommerzielle Anfragen:** <https://voxislive.com/licensing>
 
 *Voxis Live — Echtzeit-Sprachübersetzung, simultan.*