OpenMontage setup with Ollama and Google Antigravity

This tutorial builds a local AI video workflow around OpenMontage, Ollama and Google Antigravity. The presenter describes OpenMontage as an open source system that researches a topic, drafts a script, retrieves motion clips and assembles a timeline. A reference video supplies pacing and style cues. The proposed workflow includes a sample and user approval before final production.

The setup covers installing Antigravity and Ollama, downloading a model, loading Python dependencies from requirements.txt and preparing browser support. An app.py file creates a Streamlit dashboard where users enter an installed Ollama model and a video concept. A separate pipeline.py file connects the prompt and reference clip to OpenMontage, saving a production plan and an MP4 output.

The demonstration uses "neural navigator" as its prompt and a clip from an earlier video as its reference. The presenter names Gemma 4 2B during setup, but later calls the selected model Llama 4 2 B, so the model identification is inconsistent. He describes the result as a dummy video and suggests larger models or API services for better output. His zero-cost claim concerns the local pipeline; live web research and media retrieval mean the described workflow is not fully offline.