Custom Character Creation

NavTalk now supports custom character creation! Users can upload their own images to create personalized digital avatars for real-time interactions. You can train your own avatar on this page and interact with it: https://console.navtalk.ai/#/avatar/avatar.

Video Requirements

To ensure the best synthesis results, we recommend uploading a video that is at least 2 seconds long. The system will automatically capture and use the first 2 seconds of the video for processing, and the subsequent parts will not affect the synthesis results.

Video Suggestions

• Stable footage with natural expressions：Please face the camera directly, keep your mouth closed, blink naturally, and avoid speaking, large head movements, or other expression changes.

• Recommended loop effect：If the video can naturally connect from beginning to end (as shown in the Playground example), the synthesis effect will be smoother.

• Pay attention to facial features：The current model's processing of complex facial details such as beards is still being optimized, and blurring or distortion may occur. It is recommended to use portraits with little or no beard.

• Use real person videos：Real person videos help capture subtle expressions and movements more accurately, significantly improving synthesis quality.

Production Suggestions

You can first use various text-to-image AIs to generate ideal character images, then use MuseV to generate video clips that meet the above requirements, and upload them directly for more controllable and high-quality input sources.