OpenField: Open-Source Studio for Video, Stills, and 3D on Local GPUs via ComfyUI

OpenField is an open-source creative studio application designed to run locally on a user's GPU by interfacing with ComfyUI. Built for Comfy's open-source Higgsfield competition and distributed under the MIT license, the platform uses open-weight models and the official Comfy SDK to provide modular tools for video generation, photography, and 3D rendering.

Core Capabilities and Features

• Video and Stills: Supports text-to-video or image-to-video pipelines, motion presets, and multi-reference image conditioning on models like FLUX.2.
• 3D and Object Workflows: Converts single still images into textured 3D GLB models featuring PBR maps and turntable views.
• Stage and Director Tools: Allows users to block out scenes in a 3D browser interface using basic shapes and models, compiling multi-shot sequences through local language model planners.
• Model Integration: Supports local LoRAs, custom node packs, and MCP (Model Context Protocol) integration to let AI clients direct and execute tasks.

Technical Setup and Requirements

• Execution Environment: Runs locally via a connected ComfyUI (version 0.38+) instance, utilizing a Python proxy and the Comfy SDK.
• Dependencies: Requires local model downloads, specific environment variables (such as a Hugging Face read token for gated weights like LTX-2.5), and third-party tools like ffmpeg.
• Licensing: The software code is MIT-licensed, while individual model weights adhere to their respective publishers' terms.

Verification and Practical Implications

Users can verify installation requirements and view code implementations directly within the official GitHub repository (`wexloops/openfield`). The platform operates entirely locally without requiring paid cloud subscriptions or remote credits, though performance depends heavily on local hardware capabilities.