Pantelis Kalogiros · 2026-08-23 · notable
Varkos — a local AI companion that talks while it plays Skyrim with you
Varkos is a voice-driven Skyrim companion that listens, plans and acts in the game while it talks. Speech-to-text lands in 40-80 ms and a spoken reply can arrive in under 500 ms, all from local models.
A demon-dog companion that hears you, acts in the world, and answers out loud in under half a second — all on local models.
What is it?
Varkos is an AI character that plays Skyrim alongside you instead of waiting for commands. It keeps an always-on microphone open, tracks the game world, holds persistent plans, and speaks back in a voice. Pantelis Kalogiros built it to test whether a companion can feel present rather than turn-based, and the whole pipeline runs on local models.
How does it work?
Speech reaches Varkos through an optimized Qwen3-ASR 1.7b with a custom streaming harness, and replies come back through PocketTTS-Raven and an optimized Qwen-3-TTS. Between them sits ALE, an Action Latent Encoder that mixes embeddings, small classifiers, explicit rules and classic ML to decide what the character does. The game runs on Windows while audio and inference run on an M4 MacBook with 12 GB or more of GPU memory.
Why does it matter?
The latency budget is the point: 40-80 ms for voice-to-text, 20-60 ms for audio generation, about 20 ms for action analysis, and 300-600 ms to compose a reply, so a full spoken answer can land under 500 ms in the best case. That is the range where a game companion stops feeling like a chatbot. Kalogiros has already open-sourced PocketTTS-Raven and says the Qwen3-ASR harness is coming.
Who is it for?
game developers, voice-agent builders
Try it
https://github.com/pkalogiros/pocket-tts-raven