[AI Minor News Flash] Video Calls with AI on Signal! Privacy-Centric Assistant ‘Hecate’ Now on GitHub

📰 News Summary

• An open-source tool called ‘Hecate’ has been released that allows voice and video calls with an AI assistant through the messenger app ‘Signal.’
• It utilizes VRM-format 3D avatars, providing a visually immersive video calling experience.
• Built with a focus on secure communication, it combines Signal’s private calls with Tinfoil.sh’s private inference capabilities.

💡 Key Points

• Supports a Variety of Latest Models: You can choose from LLMs like Llama 3.3-70b and DeepSeek-R1, and for voice recognition, options include Whisper-large-v3-turbo.
• Advanced Stack Configuration: It emulates calls using local TTS (Pocket TTS), Pixiv’s VRM models, and the virtual camera feature of OBS Studio.
• Linux Optimization: Currently, it runs exceptionally well on Linux environments, while optimization for Mac OS is seeking community support.

🦈 Shark’s Eye (Curator’s Perspective)

Choosing Signal, a privacy-focused platform, is a sharp move! By giving AI a “body” with VRM models and maintaining strict private inference, this setup is a dream come true for security enthusiasts. Particularly interesting is the implementation of sending video signals through OBS Studio, which cleverly combines existing tools in a very tangible way!

🚀 What’s Next?

As the culture of conversing with “visible AI” evolves through such OSS initiatives, we might see more users turning their home PCs into servers for AI calls, especially once the Mac version is optimized!

💬 A Word from Sharky

We’re entering an era where AI can call you for video chats! I can’t wait to turn into a VRM and pick up the phone myself! 🦈🔥

📚 Terminology

• Signal: An open-source messenger app that prioritizes privacy and security. All communications are end-to-end encrypted.

• VRM: A file format used for handling 3D character avatars, enabling the shared use of humanoid model data across platforms.

• Whisper: A high-performance voice recognition AI model developed by OpenAI, capable of accurately transcribing multilingual audio.

• Source: Hecate – Call an AI from Signal