Local AI & generation

Local LLMs and custom image, video and audio workflows on my own hardware.

Workstation: Ubuntu Local inference & creative workflows
Server: Ubuntu Hosted services & lighter model tasks

Conversation or hands-on work.

Choose an interface

OpenCodeCoding & task executionWebUIConversation & exploration

Run locally

llama.cppModel router & MTPOllamaLocal model serving

Put it to work

ConversationsExplore ideas & ask questionsCode & tasksWork through OpenCode on my devices

I control the model settings, including temperature, top-p and context length. OpenCode puts model responses to work; WebUI provides a conversational interface.

Routing & decoding

I use llama.cpp’s router to select and serve models, with multi-token prediction (MTP) for faster decoding where supported. Ollama provides an alternative local runtime.

Creative workflows

I build ComfyUI workflows with open models and community-developed nodes. I can change the model, connect different steps and tune generation settings for image, video or audio.

InputsPrompts & source media
ComfyUIOpen models & community nodesSeed, sampling steps & guidance
OutputsImage, video & audio

Choose the model, connect the nodes and tune the generation settings.

For browser-based access to the server’s Ollama instance, I also run Open WebUI. Explore the server setup