Naia

Download

Download Naia and start chatting with your AI avatar.

v0.1.5 Windows build verified (exe + MSI). Linux Shell builds coming soon.
Windows (x86_64)
v0.1.5NEW

Ready to run after install. Naia Shell automatically installs and starts the AI backend (Gateway). v0.1.5 ships Alpha Memory System v1, Knowledge Graph, and encrypted backup.

  • naia-0.9-omni-24g runs tools during voice conversation — ask by voice and it actually creates files, searches, and runs skills, not just talks
  • Redesigned reference-audio (voice cloning) settings — card UI with in-app recording and instant preview
  • Prosody tags from voice replies map to avatar expressions and chat emoji — the avatar reacts as Naia speaks
  • Gemini Live voice UX — context bridge, barge-in (interrupt mid-speech), text/voice parity, and voice panel tools

Windows 10 / 11 x86_64. Administrator rights are requested during install. Microsoft Store registration is currently pending.

Naia OS (Live USB / Install)
Recommended

The complete Naia experience. Includes AI avatar, Korean input, and pre-installed apps. Boot from USB to try, or install to hard drive.

Live USB resets on reboot. Install for permanent use.

Shell UI only — AI features require Naia OS or separate Gateway setup.

Flatpak

Sandboxed environment. Requires GNOME Platform runtime.

Recommended for Linux — full app (login, chat, voice).

flatpak remote-add --if-not-exists flathub https://flathub.org/repo/flathub.flatpakrepo
flatpak install -y flathub org.gnome.Platform//50
flatpak install --user ./Naia-Shell-x86_64.flatpak
Flatpak

System Requirements

  • Windows 10/11 x86_64
  • Linux x86_64 (Naia OS ISO / other builds coming soon)

Verify Checksums

To verify the integrity of downloaded files:

sha256sum -c SHA256SUMS

Release Notes

v0.1.5 (2026-06-05)

Naia-0.9-Omni-24g is here — a smart AI that clones your voice, talks in real time, and runs skills, all on a single RTX 3090 PC. This release adds voice tool execution, a redesigned voice-cloning UI with record & preview, natural barge-in, 30-language support, native Ollama local AI, and MS Store & Steam distribution.

  • feat(voice)naia-0.9-omni-24g runs tools during voice conversation — ask by voice and it actually creates files, searches, and runs skills, not just talks #352
  • feat(voice)Redesigned reference-audio (voice cloning) settings — card UI with in-app recording and instant preview #349
  • feat(voice)Prosody tags from voice replies map to avatar expressions and chat emoji — the avatar reacts as Naia speaks #350
  • feat(voice)Gemini Live voice UX — context bridge, barge-in (interrupt mid-speech), text/voice parity, and voice panel tools #313
  • feat(voice)Reference voice for Naia Local — direct ref-audio, default voice selection, and mid-session voice switching
  • fix(voice)Real-time voice stability — server-VAD passthrough, transcript handling, cold-start socket races, and reference-audio auth fixes #219
  • feat(agent)Native Ollama provider — reasoning, tool calls, and num_ctx handling for fully local LLMs #357
  • feat(dist)MS Store (Win32 listing) and Steam portable distribution with conditional code signing #314
  • feat(shell)Panel spec, docs, and examples for the extensible panel system, plus a zip-tab gate #358
  • feat(launch)YouTube BGM background fallback on launch — music keeps playing even before AI connects
  • feat(auth)Rust localhost HTTP callback server — more reliable desktop OAuth/deep-link sign-in #341
  • fix(shell)0.1.5 launch bugfixes — startup, config, and provider routing corrections #342
  • fix(env)Separate dev/prod gateway environments — tauri:dev and tauri:prod target the right gateway #333

v0.1.4 (2026-05-22)

Naia Alpha — first public release. AI OS desktop app with workspace, browser panel, voice AI, memory system, and 14-language support. Windows installer available.

  • feat(shell)Workspace panel — session dashboard, file explorer, and code editor #99
  • feat(workspace)PTY terminal tabs with xterm.js — run CLI agents like Codex or Claude Code inside Naia #119
  • feat(workspace)Image, CSV, log file viewer and HWP/HWPX viewer via sidecar #116
  • feat(workspace)Naia ADK integration — skill discovery, skill launcher, session templates, workspace discovery #227
  • feat(shell)Editor reload, copy path, input history, and file navigation #194
  • feat(shell)Browser panel — Chrome/WebView2 embed with CDP tools, voice tools, and theme support #95
  • feat(browser)Browser bookmarks, shortcuts, AI toggle, and fullscreen avatar layer #263
  • feat(panels)Dynamic iframe rendering and panel API for extensible app system #89
  • feat(voice)MiniCPM-o 4.5 vLLM-omni integration — real-time voice conversation with audio reference (voice cloning) #216
  • feat(voice)OpenAI Realtime API protocol support — multi-turn flow stability and session management #219
  • feat(shell)STT/TTS provider registry — Web Speech API, Browser TTS, vLLM, and audio device selection #51
  • feat(agent)Alpha Memory System v1 — 4-store architecture with Ebbinghaus decay, Hebbian association, and consolidation #145
  • feat(agent)Knowledge Graph — entity-relationship graph with TF-IDF indexing and Louvain community detection #173
  • feat(agent)AES-256-GCM encrypted memory backup with PBKDF2-SHA256 key derivation #211
  • feat(agent)Multi-embedding support — Solar, Qwen3, BGE-M3, Gemini — with benchmark framework #189
  • refactor(agent)Hardened SQLite memory engine v6 — async worker-thread isolation for UI responsiveness
  • feat(agent)21 built-in skills from @naia-adk/skills-builtin — weather, browser, workspace tools and more #273
  • feat(agent)Agent-browser skill and welcome onboarding skill #274
  • feat(agent)MCP skill bridge — Model Context Protocol client/server integration #200
  • feat(agent)Tool safety metadata — concurrency and destructive declarations for safe tool execution #183
  • feat(agent)Token budget warning — pre-flight context overflow prevention #185
  • feat(agent)@ mention inline file/folder autocomplete in chat #195
  • feat(auth)Embedded browser login with Chrome/WebView2 — OAuth flow inside Naia window #197
  • feat(auth)Deep link authentication fallback for Windows desktop app #233
  • feat(shell)BGM player — YouTube Music integration with categories, search, and AI voice commands #303
  • feat(shell)Windows IME support — Korean 한/영 toggle and diagnostics panel #241
  • feat(i18n)14-language UI support — Korean, English, Japanese, Chinese, French, German, Russian, Spanish, Arabic, Hindi, Bengali, Portuguese, Indonesian, Vietnamese #312
  • feat(shell)Fullscreen avatar layer with VRM model loading and camera controls #253
  • feat(dist)Naia OS persistent USB — ISO integration with partitioned personal storage #262
  • fix(shell)Windows installer bundles all required DLLs (Vosk, MSVC runtime, MinGW) — fixes launch crash
  • ci(release)CI hardening — explicit DLL download, installer content verification, smoke test before release
  • refactor(agent)Consolidated AI runtime — unified gateway architecture with Naia Agent as default #201