This update adds two fully offline AI creation features: ■ NEW: AI Music Generation Describe the music you want and generate stereo audio right on your iPhone, offline. Adjust length, steps, guidance and negative prompt; built-in player (play/pause/scrub), then save and share. (Requires iOS 17+, e.g. iPhone 15 Pro.) ■ Enhanced: AI Image Generation On-device Stable Diffusion, fully offline. Switchable models, style LoRAs, reference images, automatic face fix. ■ Both • Prompts support Chinese/Japanese/Korean/Thai/Vietnamese/French and more, auto-translated to English before generating • Stability, memory and audio-quality improvements • iOS 27 support

This update focuses on quality and stability: • Sharper translations — more faithful wording, fewer dropped details and less over-editing • Smoother performance and lower memory use • Various bug fixes and refinements Thanks for using Kanon AI — your feedback is always welcome!

Fixed IAP problem

This update focuses on quality and stability: • Sharper translations — more faithful wording, fewer dropped details and less over-editing • Smoother performance and lower memory use • Various bug fixes and refinements Thanks for using Kanon AI — your feedback is always welcome!

v4.6 - Major Update AI Tools Free Daily Trial The 48 Pro tools are now open for free trial - 10 free uses per day, shared across all tools. Explore writing, business, travel, learning, creativity, lifestyle, office, and language tools across 8 categories. Hit the limit? It resets automatically at midnight local time. Translation Quality Upgrade Rewritten translation rules: the model now reads the whole passage first to grasp context, then translates faithfully - no over-polishing, no dropped phrases, no merging "redundant-looking" sentences. Closer to the original meaning. Faster Launch New "Preload AI Model at Launch" toggle. When off, the model loads only on first use, saving ~3-6 GB of memory and battery at launch. Significantly better experience on older devices (iPad mini 6, iPhone 12, etc.). One-Tap Memory Release A new "Unload AI Model" button on the Hub lets you release memory whenever you want. The model reloads automatically on next use. All AI features run 100% on your device. No cloud, no data sent. Your conversations, creations, and privacy stay with you. If you enjoy this update, please leave us a five-star review on the App Store so more people can discover Kanon AI

Kanon AI v4.5 — Small polish [Stability] - Fixed a rare crash that could occur when handling certain translation results - Clearer alert when an AI model file fails to load (no more silent file removal) [App Open Ad Fix] - Fixed: App Open ads (the launch ad) sometimes didn't appear on iOS 13+ due to a scene-lifecycle issue - Display interval tuned to 5 minutes (down from 30) to match actual app-switching patterns [iOS Compliance] - Added the "Allow Tracking" permission prompt as required by iOS guidelines — you decide whether to share tracking data with ad networks - Declining tracking does NOT affect any feature; AI translation, chat, and tutor still run 100% on-device Privacy unchanged: local inference, speech recognition, and draft storage all run on-device.

Kanon AI v4.4 — iPad Lands, AI Engine Tuned [Full iPad Support] (NEW) - iPad mini 6, iPad Air, iPad Pro — the whole lineup works now - App is Universal — no more cramped iPhone-stretched-on-iPad look - Sidebar menu redesigned for iPad layout; items no longer cut off [AI Engine Major Tune-up] - 4GB devices (iPad mini 6 / iPhone 13 / iPhone 13 mini / iPhone SE 3 / iPad) get a model upgrade: Qwen 2.5 1.5B → Qwen 3.5 2B - AI Chat no longer falls into repetition loops on smaller models (systemic fix) - Translation no longer leaks the model's English "thinking" stream (Qwen 3 native thinking disabled by default) - Sampler fully re-tuned: Translation / Chat / Tutor / Tools each get their own optimised params - Each model tier (small / medium / large) gets matching sampler settings — small models no longer punished by large-model prompts [Advanced AI Settings] (NEW — Power user) - Settings → Advanced AI Settings — tweak the sampler per task yourself - 4 tasks × 4 parameters (Temperature / Top P / Top K / Repeat Penalty) = 16 knobs - Live preview — modified values show a "(default)" indicator so you can see what was changed - One-tap reset to the default matrix [Translation Output Cleaner] - Streaming UI no longer leaks the model's English "self-talk" reasoning - Final result is clean too — no more "Sure!" "Here's the translation:" prefixes - Script-based filter for: Traditional Chinese / Simplified Chinese / Japanese / Korean / Russian / Thai / Arabic / Hindi targets [Model Management Fixes] - When a model file fails to load, you now see a clear alert instead of the file vanishing silently - Auto-detect common mistakes like "you downloaded the mmproj (vision projector) instead of the main model" — gives a specific hint - AI Chat hides photo / mic buttons on models that don't support vision [Privacy unchanged] All AI inference still runs 100% on-device. Nothing is uploaded.

Kanon AI v4.3 — Your Personal Assistant Goes Real, Your Companion Goes Personal [Personal Assistant x iOS Calendar] (NEW) - Tell the AI about meetings, birthdays, reminders, travel dates — they go straight into your iOS Calendar - Get system notifications when the time arrives, even when the app is closed - Deleting a note also removes the matching calendar event [Companion: Build a Character from Your Photo] (NEW) - Pick a photo from your album — the subject is lifted out automatically with a transparent background - Group photos, side profiles, and tilted-head shots all work - Your character blinks and breathes, just like the Live2D ones - Fully on-device — your photos never leave the iPhone [Major Translation Quality Boost] - Japanese: kanji false friends, keigo levels, loanwords, sentence-final particles all handled with more care - Korean: honorific levels, sentence-final endings, 우리 usage all properly matched - Traditional Chinese output enforces Taiwan-style vocabulary (no more Mainland terms slipping in) - Defaults to mirroring the source register; flip "Force Polite Tone" to override [Language Tutor Correction System Rewrite] - Three clear labels: "your version", "suggested", "common in casual speech" - No more nitpicking on omitted subjects, casual contractions, or common slang [Other Improvements] - Auto-saved drafts across 6 features (Translator / Dictionary / Realtime Chat / AI Chat / Companion / Tutor) - Mute toggle added to Language Tutor - Various TTS and image-parsing stability fixes Privacy unchanged: local inference, photo subject extraction, and draft storage all run on-device.

Kanon AI — Introducing Virtual Companion! Virtual Companion (brand new) • Create your own AI character — fully customize name, gender, age, personality, and description • 10 personality tags: gentle, playful, rational, humorous, knowledgeable, shy, cheerful, caring, mature, mischievous • Pick the character's reply language (8 languages) — the companion stays in character regardless of what language you speak • Multi-character: create as many companions as you like, each with independent chat history and memories • Swipe to edit or delete any time Live2D Character Animation • 4 free preset characters: Haru, Hiyori, Mark, Natori • True Live2D rendering: breathing, blinking, head sway, body sway • 4-state machine: Idle / Listening / Thinking / Talking • Transparent overlay: the character floats above the chat, messages scroll through • Procedural lip-sync: mouth opens and closes naturally while speaking Voice-first Chat • Push-to-Talk: hold the mic to speak, release to send • Apple Speech real-time recognition • AI reply auto-plays with TTS voice • Pressing the mic interrupts any TTS playback for seamless back-and-forth Memory System • Companions automatically remember what you said — background extraction every 6 messages • Memory categories: preferences, facts, life events, emotions • Memories persist across sessions — your companion still knows you tomorrow • Auto-dedupe: similar memories merge and gain importance • Auto-prune: capped at 30 per companion, frequently used memories stay • View, edit, or delete memories from the chat menu Hub Companion Shortcut • Once you have a companion, the Hub greeting shows a live animation of your first character • Tap it to jump straight into chat • Real-time rendering — breathing, blinking, head sway all animate Other improvements • LLM output cleanup: filters malformed Gemma tokens like stray • All 9 locales fully localized for companion strings • Companion chat respects the global Thinking Mode setting • Clearing chat history now also clears that companion's memories Privacy • All companion conversations and memories stay 100% on-device in local SQLite • Memory extraction runs on the local LLM — nothing ever leaves your phone • Uninstalling the app wipes all data

Kanon AI — Multimodal AI Upgrade! AI Image Understanding (New) • Upload photos in AI Chat — Gemma 4 analyzes images locally • Camera or photo library, fully offline • Translate menus, signs, and documents by photo AI Audio Understanding • Record and send voice in AI Chat — Gemma 4 understands speech directly • Up to 30 seconds, 16kHz quality • No text = voice conversation; with text = custom instructions Camera Translate: 3 Modes • Standard (OCR + Translation) • Cloud AI (Gemini Vision) • New "Local AI" — offline image translation Notebook (New) • Save any AI result for later review • Search, rename, and browse by source Language Tutor (New) • Free-form language learning with session management AI Tool Voice Playback + Custom Models • Play results aloud with auto language detection • Enter your own GGUF model URL Improvements • 3-layer hybrid language detection (Chinese/Japanese accuracy) • AI Chat streaming performance optimized • Voice input: review text before translating • TTS auto-retry, numerous bug fixes

Kanon AI is here! New Direction: Offline AI Assistant • Core experience shifted from cloud translation to on-device AI • Everything except cloud voice runs on your iPhone — no data uploaded 4 On-Device AI Models • Lite / Light / Smart / Pro auto-recommended based on iPhone RAM • Powered by Gemma 4 and Qwen 2.5 series • Translation, dictionary, and chat all work offline 50+ Kanon AI Helpers • Writing polish, rewriting, grammar check, email drafting • Brainstorming, travel phrases, learning aids • Full support for 17 languages AI Chat (Offline) • Smooth multi-turn conversations, running entirely on iPhone • Chat history auto-saved for review anytime • No conversation data ever leaves your device New Hub Home • One-stop entry for every AI tool • Quick access to favorites and recent items High-Quality Voices • 80+ languages, 400+ neural voices • Assign dedicated voice characters per language

Major Update - New Features! 【Chat Room】Real-time Messaging • Chat with friends with instant translation • Automatic text message translation • Send image messages • Push notifications for offline messages • Cloud-synced chat history 【Camera Translation】Photo Translation • Take photos or select from album • AI-powered text recognition • One-tap translation to target language 【Quick Settings】Easy Access • Switch source/target languages instantly • Quick access to common settings • Available on all feature pages 【Other Improvements】 • Improved translation speed and accuracy • Optimized user interface • Bug fixes

What's New in This Version Brand New Real-Time Chat Connect two iOS devices directly for one-on-one translated conversations! No internet required – establish secure peer-to-peer communication via Bluetooth and Wi-Fi, making face-to-face cross-language conversations seamless and natural. Premium AI Voice for Subscribers Subscribers and VIP users now enjoy AI-powered text-to-speech with superior voice quality. Experience more natural intonation and accurate pronunciation that dramatically enhances your translation listening experience. Smart Language Detection Introducing smart mode – the system automatically detects translation direction without manual language switching. Please note: Voice input requires explicit language selection even in auto-detect mode. Enhanced Conversation Features Completely redesigned conversation interface now displays full conversation history on screen, allowing you to review previous translations anytime for more coherent communication. Automatic Audio Playback When using headphones with Conversation or Real-Time Chat features, the app automatically plays translated audio, letting you focus on listening without missing important information.

Latest Version Improvements: • Enhanced app stability • Voice playback after translation changed from automatic to manual user control

Fix the issue where subscription restoration fails

Updates: 1. Integration of three AI translation engines 2. New dictionary feature 3. New subscription mechanism 4. Switched to TTS audio playback