Changelog

    What's new in ShadowTalk AI

    Release Notes

    Every improvement, documented

    Workspace-first entry, persistent auth, App Builder, marketplace agents, and hardware-aware routing — plus admin-published updates from the CMS.

    v3.4.0
    November 13, 2026
    Latest

    LIVE Mode: A Voice Presence, Not a Chatbot

    Complete redesign of live voice conversation: an animated living avatar, hands-free auto barge-in, streaming sentence-by-sentence speech, and a natural speech rhythm engine.

    LiveAvatar: a canvas-rendered living presence that breathes, blinks, drifts with micro-saccades, morphs with real microphone/TTS amplitude, and re-tints its aura per phase (idle/listening/thinking/speaking) — replacing the static CSS orb.
    New
    Hands-free barge-in: the mic keeps metering (echo-cancelled) while the AI speaks; sustained speech interrupts it — talk over the AI mid-sentence and it hands you the floor. Tap-to-interrupt still works.
    New
    Streaming speech pipeline: replies are spoken sentence by sentence as the LLM generates them (first audio ~1s after you stop talking) instead of waiting for the entire response.
    New
    Speech rhythm engine: punctuation-aware pauses (150–420ms), abbreviations-safe sentence splitting, and emoji/markdown/URL stripped before TTS so nothing robotic is read aloud.
    New
    Immersive full-screen live UI: ambient stage, live captions for both sides of the conversation, compact transcript strip, keyboard fallback input, and mobile safe-area padding.
    Improved
    Voice-natural persona rules for spoken replies: no markdown or emoji, varied sentence rhythm, react-then-answer flow.
    Improved
    Barge-in is guarded by an echo grace period and cooldown so the AI never interrupts itself.
    Security
    Voice
    Avatar
    Barge-in
    Streaming
    v3.3.0
    November 13, 2026

    The Intelligent Workspace: Stealth RAG, Token Optimizer & MCP Bridge

    The workspace becomes intelligent: local-first document retrieval, provider-aware context budgeting, approval-gated local tools — plus production hardening for memory, chat history, sign-in, and mobile.

    8 changes — tap to expand

    v3.2.0
    October 22, 2026

    Hybrid API Key & OmniRoute Gateway Integration

    Launched the cohort-based onboarding system (Instant Start vs BYOK) with OmniRoute gateway routing for non-tech users. Added real-time /status page and /security trust center.

    6 changes — tap to expand

    v3.1.0
    September 15, 2026

    PWA Offline Support & 11-Locale i18n

    Added Progressive Web App offline caching, 11 hreflang locales, and Puppeteer-based prerendering for 13 SEO-critical routes.

    5 changes — tap to expand

    v3.0.0
    August 31, 2026

    The Agentic Pivot: Less Bloat, More Power

    Integrated OpenAI GPT-4o as a first-class citizen, and stripped away 40+ legacy pages to focus strictly on Omniscience and Twin.

    4 changes — tap to expand

    v2.6.0
    June 1, 2026

    Workspace-first experience

    Open ShadowTalk straight into chat — faster entry, persistent sessions, cleaner composer.

    7 changes — tap to expand

    v2.5.0
    May 30, 2026

    App Builder & runnable Marketplace

    Generate full web/mobile projects in the IDE; install agents that actually run in chat.

    3 changes — tap to expand

    v2.4.0
    May 28, 2026

    Hardware-aware speed paths

    Automatic local WebGPU/WASM vs cloud routing on capable devices.

    4 changes — tap to expand

    v2.3.0
    May 20, 2026

    & agentic tools

    Bring your own API keys and expanded chat tooling.

    3 changes — tap to expand

    v2.2.0
    May 10, 2026

    Foundation release

    Desktop app, trust metrics, brand refresh, and an upgraded chat experience.

    4 changes — tap to expand

    Try the latest

    Open the workspace for instant chat, or read full guides on the docs page.