PrivaEmanator

PrivaEmanator brings a fully offline AI companion to your desktop. Powered by local AI — no cloud, no subscription, no data leaving your machine.

⚠️⚠️ Please note: As all AI runs locally on your hardware, response times may vary depending on your hardware specifications. Only NVIDIA RTX 30, 40, and 50 series GPUs are supported, with a minimum of 12GB VRAM required. Please consider this before purchasing.

Features

• Conversation — real-time chat powered by local LLM

• Agent — extend your companion's capabilities by adding custom skills with CLI execution support.

• Vision — powered by local VLM, sees your screen and reacts to what you're doing, Can be toggled on or off in settings (⚠️CUDA only)

• Voice output — TTS with natural voice (⚠️Chinese & English) 

• Voice input — speech-to-text support

• Load VRM 1.0— bring your own 3D character (⚠️VRM 1.0 only)

• Free camera — rotate and view from any angle

• Touch interaction — reacts when you interact

• Memory — remembers your conversations and learns about you over time

• Multiple themes — Normal, Scanline, Particle

LLM Model Support

Supports GGUF format models. Recommended: Qwen3 4B Q4 (tested). Other models including Llama, and Qwen series may work but have not been tested. Please evaluate based on your own setup. 

You can add your own system prompts, conversation histories, and memory files to the corresponding folders under the assets directory.

Agent Skill Support

Extend your companion's capabilities by adding custom skills with CLI execution support.

VRM 1.0 Support

Supports VRM and VRMA. You can add VRM characters to the assets/vrms directory.

Custom VRMA animations are also supported — create your own animation folders under assets/vrms/[character]/vrma/ and place your VRMA files inside.

TTS ONNX Support

Supports TTS ONNX voice models, You can add your voice models to the corresponding folders under the assets directory.