Meet Deepy your friendly WanGP v11 Agent. It works offline with as little of 8 GB of VRAM.

A new open-source AI agent called Deepy is making waves by offering powerful, offline AI capabilities without subscription fees. Built on the WanGP v11 architecture by developer deepbeepmeep, Deepy requires only 8GB of VRAM to run locally, making it accessible to users with modest hardware. Unlike cloud-based services from OpenAI or Anthropic, Deepy operates completely offline, ensuring user privacy by never sending data to external servers.

Deepy excels at automating tedious media tasks like generating black frames, cropping videos, extracting specific frames, and trimming audio. Its real power lies in orchestrating complex workflows across multiple AI models including LTX-2.3 for image generation, Wan for processing, and Qwen3 TTS for audio synthesis. For example, it can generate an image of a robot disco dancing on a horse, edit it to show the robot standing beside the horse, verify the edit matches the description, and create a transition between the two images—all in one automated sequence.

The agent can also create complete multimedia presentations by generating a self-portrait image, producing an audio introduction describing its capabilities, and combining them into a video. Available on GitHub, Deepy represents a significant step toward democratizing advanced AI workflows, putting sophisticated multi-model automation within reach of individual developers and creators who value privacy and cost-effectiveness.