Version 3.1.0

LynxCatAI support
- Visual Novel Mode, character focus, story highlighting, and local narration now work on LynxCatAI.
- VN uses native character and location cards without requiring an import. Extension pictures take priority, with website pictures as a fallback.
- Card import and sync carry names, types, triggers, gender, and website image links while preserving local media.
- A LynxCatAI-specific VN theme matches the site's dark panels and blue-purple accents. AI Dungeon keeps its existing style.
- The normal-story narrator is available directly inside played LynxCatAI stories. It stops when you leave the story or enter VN.
- Native turn entry, Continue, and Retry are supported. AI Dungeon-only retry controls are hidden. TTS initialization on LynxCatAI is fixed.

VN writing instructions
- Track Speakers now uses Author's Note in AI Dungeon, adding the component when missing, instead of creating a VN Mode story card.
- On LynxCatAI, it uses AI Writing Style for the played story.
- Your own notes are preserved. The marked VN block is removed when you turn Track Speakers off, exit VN, or load a fresh page. Temporarily minimizing VN keeps the session active.
- Recognized old VN instruction cards are deactivated. Open edits and full fields are protected. Dialogue formatting changes may take 2 or 3 turns to appear.

Voices and settings
- Story cards now offer Male, Female, or None for Gender. With Track Speakers on, VN selects the corresponding dialogue voice; None uses narration.
- Dialogue offers M1-M4 and F1-F4, while narration uses Male (M5) or Female (F5). Speed and pitch are saved per voice.
- VN settings are split into Narration & Voices and TTS Engine tabs.
- Larger, reorganized settings, Story cards, and Card sync screens offer clearer explanations, visible input fields and tabs, and more space for cards.

Scene fixes
- Resumed dialogue after the same speaker's action keeps the correct speaker, while the action remains narration.
- Automatic backgrounds follow names and triggers on Location cards, including dialogue mentions. The last matching location stays until another is mentioned. Character and Other cards are ignored by location tracking.

Version 3.0.1

AI Dungeon compatibility
- Fixed duplicated action text such as "Action You say ... You say ..." caused by new hidden accessibility labels, confirmed on Alpha. These labels no longer appear in story text or VN narration. Actions were not being submitted twice.
- Restored latest-response detection and prevented hidden labels from affecting text styling. Regression checks cover Say, Do, Story, and Guide.

Visual Novel Mode
- Renamed Story-card instructions to Track Speakers. The help explains that changes to the AI's writing may take 2 or 3 turns to take effect.
- Turning Track Speakers off immediately disables speaker highlighting, foreground emphasis, and character dimming, including in dialogue that already has speaker labels.
- Simplified settings help and narration status messages. Accelerated TTS help now shows the Windows helper and extra-tab requirements only for Firefox.

Reading and scene controls
- Auto advances with a reading pause based on word count and waits for active narration to finish. Hover over Auto for Auto Continue, which requests the next AI response using your selected model and credits.
- Hide/Show lets you enjoy the full scene. Top controls fade when idle and return on hover or keyboard focus.
- Location choices are remembered per adventure, including Automatic and No background. Missing location cards fall back to Automatic.
- Improved portrait spacing on large displays, including 4K.

Dialogue and character fixes
- Track Speakers also recognizes explicit speech attributions such as "Careful," you whisper. These receive the same speaker glow, foreground placement, and listener dimming as named dialogue.
- Characters remain visible through connected dialogue and narration. Names mentioned only inside quotes do not introduce absent characters.
- Titles such as Dr., Mr., and Mrs. no longer split names prematurely. Separate speakers' dialogue stays separate.

A hidden narration extra
- A discoverable narrator reads the normal story view, with autoplay for new passages and visible Play/Pause, Stop, and Exit controls.
- Ten named voices, Male/Female 1-5, with a separate remembered voice choice. VN retains its Male/Female options.
- Entering VN exits the hidden narrator. Update the Firefox TTS Helper to use the additional voices with acceleration.

🎭 Visual Novel Mode
Read with character portraits, location backgrounds, and speaker focus. Structured dialogue brings the speaker forward and dims the rest; mentions inside dialogue do not summon absent characters. Track locations automatically or choose one manually. Disable blur, glow, or story-card instructions for a lighter mode.

Use Do/Say/Story/Guide, Story Card @mentions, Retry with instructions, and retry history inside VN. Reading positions are saved; click the page counter to jump, or press Space for Next. Right-click Exit VN to temporarily return to the story. Audio and action panels appear on hover; narration settings are in the top Settings button.

🔊 Local narration
Choose Male/Female, 5-10 generation steps, pitch in half-semitone increments, and a queue of 1-20 upcoming lines. Preview your voice settings, initialize cached models, and see how many lines are ready. Dialogue labels are not narrated. Text waits for audio, with Read now available to skip buffering.

Optional acceleration offers 2/4/6 threads, defaulting to 2. Firefox uses the optional Windows x64 TTS Helper and a grouped engine tab. Download its setup wizard from Firefox setup in settings. No administrator rights, Node, or separate .NET installation are needed. Firefox starts and stops the helper automatically. Standard single-thread narration needs no helper. Model downloads use the extension's background helper; speech is generated locally.

🦊 Simpler controls and fixes
Four floating quick actions, an optional compact corner layout, and session-only hiding. Improved player-action detection, reading-position recovery, dialogue grouping, scenario headings, and safe highlighting of names such as Aria. Diagnostics include TTS status.

Story Card autocomplete supports configurable card types. GitHub card sets can check for updates with merge/overwrite choices; exports offer version selection. Deleted Trinetra images can be selected again immediately.

🔒 Security and permissions
Updated DOMPurify, devalue, PostCSS, and nanoid to patched versions. New Firefox permissions: nativeMessaging for the optional local helper and tabGroups for the story/engine group. Model-host access enables voice downloads. Narration text stays on your computer.

🩹 Player actions keep their text size
Fixed player actions appearing much smaller than the surrounding story when AI Dungeon applies text styling inside nested elements. The extension now preserves the original font size, font family, line spacing and letter spacing when rendering these actions.

🔓 No new permissions.

🧭 The floating button grows a ring
Hover the floating button and four quick actions appear around it. Sets switches, creates or imports a card set and Stamp binds the set to the story you are playing, both right there without opening the editor. AID Sync and Settings open the editor on that tab; the button itself opens it where you last left it. Hovering a ring button opens its second level, clicking pins it. The ring uses whatever room the button has: a full cross in the open, a half ring on an edge, a quarter in a corner. From feedback in the AI Dungeon Discord, thank you.

🗑 The Editor entry in AI Dungeon's menu is gone
It was cloned into AI Dungeon's own menu and broke whenever that menu changed. The floating button, its ring and the keyboard shortcut cover everything it did.

🔓 No new permissions.

🎯 Stamp once per scenario, not once per adventure
A card set can now follow the scenario an adventure was started from. Stamp it once and it loads for every adventure you start or duplicate from that scenario, so restarting a story no longer means picking the set and stamping it again. Sets imported from AI Dungeon get this link automatically. The per-adventure stamp is still there as "This adventure only" and wins over the scenario link. Importing into a duplicate of an adventure you already imported now merges instead of creating a second set.

🧭 Shortcuts on the floating button
Hover the floating button and three shortcuts slide out: Story Cards, AID Sync and Settings. Each opens the editor straight on that tab, so bringing cards in from AI Dungeon is one click. Turn them off under Settings, Floating Button, Quick Actions if you prefer the plain button. The button now wears the extension icon, its size is a slider, and a Toggle Shortcut (Ctrl+Shift+F by default, recorded inside the extension) shows or hides it mid-story.

🩺 Diagnostics
The report now says whether a set is linked by adventure or by scenario.

Both ideas came from feedback in the AI Dungeon Discord. Thank you.

🩹 Fix
Player actions containing markdown came out garbled after AI Dungeon started labelling them for screen readers with a copy of the text. Those labels are now stripped before formatting.

🙏 Credit
The editor footer now names Claudia, whose Dungeon Extension v2 this project continues, and links her repository.

🔓 No new permissions.

🔓 No new permissions
This update asks for nothing 2.0.0 did not already have, so it installs without a prompt.

🔁 Importing again keeps your cards current
Import used to skip any card it had already brought over, so once imported, a set never followed AI Dungeon again. Cards you already have are now updated in place, matched by AI Dungeon's own id or by name, and only name, type and triggers are touched. Your icons, portraits, audio and colours stay exactly as you set them.

⚡ Faster editing with big card sets
Every edit used to rewrite every adventure you have. Each adventure is now stored on its own, so a change costs one adventure, not all of them. Existing data moves over automatically the first time this version loads.

🗜 Images are compressed as you add them
New uploads are shrunk on the way in, at the icon or portrait size for the slot you add them to, so a full-resolution photo never lands in storage. On by default, with a switch under Image Compression if you want originals kept.

💾 Back up everything in one file
A "Back up first" button next to Compress downloads every adventure as one file, and Import on the Adventure tab restores it.

⌨ Keyboard shortcut
Ctrl+Shift+L opens the editor. Rebindable in Firefox's extension shortcut settings.

🖼 Trinetra: key in Settings, picker remembers your folder
The API key now lives under Settings > Extension > Trinetra, entered once. The image picker reopens in whichever folder you were last in, with a clickable path back to the root, so editing several cards from one folder no longer means clicking down the same path each time.

🔧 Under the hood
• Two of the four "unsafe innerHTML" warnings this listing showed are gone; the other two are library internals.
• Build tooling updated (WXT 0.21) with stricter type checks applied throughout.
• The core logic now has automated tests.

⚠ This update asks for new permissions
Firefox will ask you to approve access to openrouter.ai and civitai.com. Those are image generation services, and the extension only contacts the one you pick, only after you paste in your own API key. If you never turn image generation on, nothing is sent to either. Everything the add-on already did is unchanged.

🎨 New: generate card images from a prompt
Add your own OpenRouter or Civitai key and a Generate option appears wherever you add an icon or portrait. Generations are billed to your own account and the cost is shown after each one. Choose whether the result is uploaded to Trinetra, so the card stores only a link, or compressed into the card itself.

• OpenRouter offers every image model it currently has, cheapest first.
• Civitai takes a model link and looks up the rest, including the sampler, steps and CFG the model's own sample images used.

🗜 New: shrink images you have already added
Settings > Extension > Image Compression finds every image stored inside a card, shows what they cost, and re-encodes them. Linked images are untouched and anything already small enough is skipped. Worth running if the editor feels slow: an adventure with a few full-resolution photos pasted in can reach hundreds of megabytes, and all of it loads on every AI Dungeon tab.

🗂 Reorganised
• Settings is now two tabs: Extension for the add-on's own behaviour, Story Cards for what you see while playing.
• The Adventure tab groups cards by type into collapsible sections, with search across names and trigger words, and filters per type.

🔧 Fixes
• Tooltips no longer open on top of the setting they describe, which had made some fields impossible to use.
• Card tilt and shine are now adjustable, and can be turned off.

🔧 Fixed: story cards were no longer being detected
AI Dungeon changed how it loads story cards, and the extension was still looking in the old place. It quietly found nothing, so a new setup could not import anything, and with no cards there was nothing to highlight and therefore no icons and no portraits either. It now finds them wherever AI Dungeon puts them, so the next time something moves this should keep working.

If you were stuck on "No story cards detected": open your story cards in AI Dungeon once, or reload the adventure, and they will appear ready to import.

• Imported adventures are named after your story again, instead of all being called "Imported Adventure".

🩺 New: Diagnostics
Settings > Support > Diagnostics checks everything the extension depends on: whether AI Dungeon's page still looks the way it expects, whether your card images actually load, and whether any of your own settings would hide icons or portraits. One click copies a short report you can paste into a support thread. It contains counts only, never card names, story text, or image links.

🎯 New: a floating Editor button you can put anywhere
The quick-access button has left the gameplay action bar. Another extension injects into that same row, and the two kept fighting over it. In its place there is now a small floating button on top of the game:
• Click it to open the editor, drag it to park it wherever suits you.
• It remembers where you put it and stays on screen when you resize the window.
• It stays faded until you point at it, and hides itself while the editor is open.
• Switch it on or off under Settings > Interface > Floating Button, which also has a Reset Position button if it ever ends up somewhere awkward.

The Editor entry in AI Dungeon's top menu still works exactly as before, so nothing is lost if you would rather keep the floating button switched off.

⚡ New: one-click Editor access
• No more digging through the top menu. There's now a DExtV2R button right in the gameplay action bar, next to Do / Say / Story / Guide / See. One click opens the extension editor, and it's styled to soft pastel colors
since AID's theme options are hard to track.

🩹 Fixes & polish
• No more phantom highlights: starting an adventure you never added to the extension now shows a clean empty state instead of the previous adventure's card highlights.
• Version footer + update nudge: the editor always shows the current version (with credits), plus a green badge when a newer build is available on GitHub.

New
- Import story cards from AI Dungeon: a new Import tab reads the story cards from the adventure you have open, lets you filter them by type (characters, locations, factions, ...), and adds them in one click instead of recreating them by hand.
- Auto-load per adventure: imported card sets are now linked to their AI Dungeon adventure and load automatically when you play it, no more switching manually. A new "Stamp Adventure-ID for autoload" button links your older card sets too.

Fixes
- Large GitHub-hosted adventures now import reliably on Firefox.
- Story text no longer shows raw markup after recent AI Dungeon layout changes.
- Player actions now render and highlight correctly, at the right font size.