Skip to main content

​
2026

July 25
V2.3.4
2026

​
1. Meeting Mode

Meeting Mode is now available for real-time digital human conversations. It lets users create and join live meeting sessions with avatar-based interaction, participant handling, real-time responses, meeting duration controls, and billing support.

​
2. Multi-Action Avatar

Multi-action avatar behavior is now supported across realtime and meeting scenarios. Avatars can use richer action orchestration, improved action selection, multi-action video support, and more natural idle/listening motion playback.

​
3. Watermark Support

Free users will now see watermarks on realtime calls and video synthesis. Becoming a NavTalk member permanently removes these watermarks.

​
4. Realtime Stability Improvements

Realtime sessions now handle startup, reconnection, interruption, hangup cleanup, avatar loading, and playback continuity more reliably. This reduces stuck loading states and improves recovery when providers or network connections are interrupted.

​
5. Avatar and Sharing Improvements

Avatar sharing and media workflows have been refined, including editable shared avatars, shared link expiration handling, improved media download behavior, and fixes around avatar loading and copied avatar data.
June 27
V2.3.3
2026

​
1. Transparency Mode

Transparency Mode is now available for realtime digital human sessions. It lets you connect your own voice or conversation pipeline, receive user messages through MQTT, and send replies back for avatar playback.

​
2. MQTT Operator Token API

A new MQTT operator token API is available for Transparency Mode. Use POST /api/open/v1/mqtt/operator-token to obtain the broker URL, scoped MQTT token, and topic templates needed by a trusted backend or the Python operator agent.
June 12
V2.3.2
2026

​
1. Realtime Call Token API

You can now create realtime call connections through a new token-based API. This lets your backend validate access and return a ready-to-use WebSocket URL, so you can manage your API key more securely without exposing it directly in the client.

​
2. Shared Session Quota

Realtime calls, Video Compose jobs, and Avatar generation now share the session quota included in your plan. This makes capacity management easier to understand and keeps usage limits consistent across major creation workflows.

​
3. Feedback Bug Fixes

We fixed reported issues from customer feedback to improve day-to-day stability and make the platform more reliable across realtime, video, and avatar workflows.

​
4. Improved Avatar Resolution Support

Avatar generation now supports both landscape and portrait formats: 1920x1080 and 1080x1920. This fixes the previous aspect-ratio limitation and makes it easier to create avatars for different screen layouts.

​
5. New Avatar Cover API

A new Avatar Cover API makes it faster to retrieve an avatar preview image without first loading the full avatar detail response.
May 22
V2.3.1
2026

​
1. Code Snippet Experience Upgrade

We’ve redesigned the in-product code snippet component (copyCode) to streamline how you move from example to running app:
  • Cleaner UI: Refreshed layout with clearer language labels, better contrast, and a more compact toolbar.
  • One-Click Download as HTML: Save any snippet as a self-contained .html file with a single click, no manual copy-paste or wrapper boilerplate required.
  • Faster Copy Action: Improved copy feedback and keyboard accessibility for a smoother workflow.

​
2. OpenAI Realtime Model Selection

You can now select OpenAI Realtime 1.5 or Realtime 2.0 when configuring real-time conversations, allowing you to balance latency and quality based on your use case. Learn more in the OpenAI Realtime API documentation.
May 15
V2.3.0
2026

​
1. New Public Home Page Experience

We’ve launched a brand new public Home page that allows you to quickly experience NavTalk’s core features without logging in:
  • Quick Avatar Customization: Generate and customize your exclusive digital avatar through simple text descriptions
  • Instant Video Preview: Play and preview video content directly on the homepage for a what-you-see-is-what-you-get experience
  • Lower Barrier to Entry: New users can fully understand and experience product features before registration, making it easier to get started

​
2. Comprehensive Message & Notification Upgrade

To help you stay informed of system updates in a timely manner, we’ve comprehensively optimized the messaging system:
  • Real-time Message Count: Message counts update in real-time, ensuring important information is never missed
  • Optimized Message Display: Brand new message list interface with clearer information hierarchy
  • Convenient Message Operations: Support for quick detail viewing, batch mark as read, and other common operations

​
3. Sharing & Avatar List Optimization

  • Enhanced Avatar List: Optimized list display for smoother browsing and management
  • One-Click Copy: Copy avatars to quickly enter editing mode and generate new avatars

​
4. Bug Fixes & UX Improvements

This update also addresses user-reported issues:
  • Improved interface styling and interaction details for a more user-friendly operation experience
April 16
V2.2.0
2026

​
1. Avatar Training File Updates

  • Adjustments to avatar training file processing.

​
2. Video Creation & Transfer Service Updates

​
2.2 Video Creation Workflow Enhancements

  • 2.2.1 Create video by selecting an avatar
  • 2.2.2 Create video from text input with selectable audio
  • 2.2.3 Added video generation progress display
  • 2.2.4 Integrated GPT model for video script generation

​
2.3 Sharing feature Optimization

  • 2.3.1 Shared avatars can be directly called without create navtalk account.
  • 2.3.2 Added sharing controls:
    • Password protection
    • Usage count limits
    • Time duration limits
  • 2.3.3 Sharing flow improvements

​
2.4 Avatar Creation Enhancements

  • 2.4.1 Added image cropping during upload
  • 2.4.2 AI generation logic improvements:
    • Prioritize generating a 4-second loop video (first & last frame identical) from image, then proceed with avatar training
  • Added support for multiple video generation providers: Gemini-V3, Kling, Vidu (currently in use)

​
2.5 Models & Voices Management

  • Providers could return selectable models / voices. Supported on both console and local environments
Improvements
  • 2.5.1 Lazy loading for models and voices
  • 2.5.2 Voices: Pagination support, Search functionality
  • 2.5.3 Open API: Added pagination support, Maintains backward compatibility (non-paginated requests are internally paginated and returned fully)

​
2.6 Streaming Quality Optimization

  • 2.6.1 Added streaming quality priority settings at project level:
    • Configurable min / max / default bitrate
  • 2.6.2 AI service supports real-time quality switching
    • Backend ready, frontend integration pending

​
2.7 Real-time Conversation Improvements

  • Improved natural transition between conversations
  • Added default handling for smoother interaction

​
2.8 Avatar Pagination

  • Added infinite scroll / load more for avatar list

​
2.9 Cartesia Custom TTS Integration

  • Integrated text recognition capability
  • Requires updates on Cartesia side (see video)
  • New Cartesia Demo repository to host related implementation

​
2.10 ElevenLabs / Cartesia

  • ElevenLabs supports platform-based model / voice creation
  • Cartesia supports platform-based voice creation
March 14
2026
New Features
  • AvatarId Parameter Support: Added avatarId parameter for WebSocket connections, enabling precise avatar lookup by ID instead of name.
  • Duplicate Name Warning: When multiple avatars share the same name, the system automatically selects the most recently updated one and sends a warning message (type: conversation.warning) to the client.
  • Query Priority Mechanism: avatarId parameter has higher priority than name parameter when both are provided.
Enhancements
  • Avatar Query Optimization: Added ORDER BY updated_at DESC to avatar name queries to ensure consistent results.
  • Code Quality: Improved avatar validation logic to reduce code duplication and improve maintainability.

​
2025

December 20
2025
New Features
  • Unified WebSocket Connection: Merged two separate WebSocket connections into a single unified connection (wss://transfer.navtalk.ai/wss/v2/realtime-chat).
  • Intelligent Parameter Configuration: Simplified connection parameters with default value mechanism. Only license and name are required.
  • Custom Event Encapsulation: All event types now use custom event encapsulation with NavTalkMessageType constants.
Enhancements
  • Response Speed Optimization: Achieved industry-leading response latency of approximately 200ms through model and system optimization.
  • Simplified Session Configuration: Session configuration now passed via URL query parameters instead of messages.
November 6
2025
New Features
  • New Voice Styles: Added cedar and marin voice styles for real-time digital human conversations.
  • Dialect Support: Added support for dialects in real-time digital human conversations.
  • 21 New System Characters: Introduced 21 new system characters for enhanced digital human interactions.
Enhancements
  • Video Stability Fix: Fixed video instability issues during real-time digital human conversations.
  • Model Upgrade: Enhanced lip-sync accuracy and audio-visual synchronization precision for digital humans.
May 29
2025
New Features
  • Character Definition: Supports the complete construction of instructions, including voice styles, function rules, and contextual memory.
  • Voice Diversity: Introduced 9 new professional voice configurations, covering styles such as customer service, broadcasting, teaching, and entertainment comprehensively.
  • Contextual Memory Mechanism: Added dual-channel strategies with messageConfig and conversation.item.create.
  • Function Calling: Introduced a full-chain example for Function callbacks, supporting the invocation of external services like CRM, weather, and search.
EnhancementsNo enhancements were made in this release.
May 15
2025
New FeaturesNo features were added in this release.Enhancements
  • TTS Synthesis Enhancement: Supports emotional synthesis in voice, compatible with Nova, Fable, and Echo styles. The TTS interface has added voice parameter validation.
April 10
2025
New Features
  • Function Calling Officially Open for Testing: Supports defining custom functions with the tools field, enhancing the operability of digital humans. Can be applied in scenarios such as weather inquiries, knowledge retrieval, and business system integrations.
EnhancementsNo enhancements were made in this release.
March 10
2025
New FeaturesNo features were added in this release.Enhancements
  • WebRTC Video Channel Stability Optimization: Enhanced ICE connection logic, supporting automatic reconnection after disconnection. Latency control optimized to less than 500ms.