Overview
Elite Intel is an OFF LINE private LLM powered ship computer for Elite Dangerous. It understands natural language and does not require memorizing command lists. It interprets intent from spoken input. Natural phrasing is accepted. Formal command syntax is not required.
The full stack (Speech-to-Text, LLM, and Text-to-Speech) can run completely offline on local hardware. No cloud connection, subscription, or data transmission is required in offline mode. Components can also be mixed: local STT with a cloud LLM, or cloud TTS with local inference.
-
Speech Recognition Powered by NVIDIA Parakeet STT model. Runs locally on CPU. Does not require training, works out of the box
-
Text To Speech Powered by Kokoro model. Run locally, private. Provides 11 voice choises.
-
Local LLM (AI) Powered by tulu3.1:8b-Supernova runs locally, private. Provides inference for uttered commands and queries without requiring user to memorize the list of commands.
-
Cloud LLM, choose between Deepseek, Grok, Claude, OpenAI or Gemini. Not private, runs in the cloud, cost money, but off loads the inference from your GPU to the cloud. This option is for commanders who do not have the necessary hardware to run local LLM.
See GitHub project for installers, wiki.
Join Matrix community to chat with dev and testers.