Show and Tell: CastWorkshop — Open-source, fully local AI character and video creation with Ollama + ComfyUI

CastWorkshop, a free open-source application that coordinates Ollama and ComfyUI to create persistent AI characters and generate images, videos, and interactive scenes fully locally, has been released as a public beta. The tool defaults to Krea 2 and MiniMax H3 workflows, supports custom ComfyUI workflows, and adds Telegram integration, with no cloud-generation subscriptions or API fees. The developer tested CastWorkshop on an RTX 5060 Ti with 16GB VRAM and 64GB system RAM, where images typically take a few minutes and videos take considerably longer; Windows is the primary tested platform, while Linux and macOS compatibility has not been validated.

Hi everyone I recently released CastWorkshop , a free, open-source application for creating persistent AI characters and using them to generate images, videos, and interactive scenes entirely on your own hardware. The idea was to make it easier to work with local AI models using natural language, without having to manage individual prompts and workflows for every interaction. CastWorkshop acts as a coordination layer between Ollama and ComfyUI , bringing character creation, conversations, image generation, and video production into one workspace. Some of the key features: Persistent characters with personalities, appearance references, voices, and conversation histories. Multi-character scenes with two characters interacting in generated images and videos. Video continuation to extend existing generated sequences. Flexible workflows — defaults to Krea 2 and MiniMax H3, but supports custom ComfyUI workflows. Telegram integration for interacting with your characters remotely. Everything is generated locally. There are no cloud-generation subscriptions or API fees. Hardware: I’ve been developing and testing CastWorkshop on an RTX 5060 Ti with 16GB VRAM and 64GB system RAM. Generation isn’t instant—images typically take a few minutes, and videos take considerably longer. It’s currently a public beta, primarily tested on Windows. Linux and macOS compatibility haven’t been validated yet. Links: I’d love feedback from anyone interested in local AI, character consistency, or combining language and media generation workflows. I’m particularly interested in hearing about installation experiences and what people would want to build with it. Your Cast. Your Stories. Your Machine.