OpenMontage setup: video production with coding assistants

OpenMontage turns a coding assistant into a video production workflow. The presentation describes an agent that researches a topic, writes a script, sources visuals, adds narration and music, and renders the result. Users approve the proposed direction before asset production starts.

The setup walkthrough begins with cloning the repository and running make setup. That command installs Python packages, the Remotion project, and Piper TTS. Users then open the folder in their assistant, which reads the guide, runs preflight checks, and presents the available production options. The speaker says the initial workflow needs no API keys; Piper provides free, offline speech, and Remotion renders locally. Optional keys enable paid voices and video generation services.

A reference workflow accepts a YouTube Short, Reel, or TikTok. The agent examines its transcript and pacing, proposes two or three concepts, and waits for a selection. For documentary montage, it searches footage from public sources by meaning through CLIP embeddings. Generated motion uses separate video providers.

The presentation describes 11 pipelines stored as editable YAML manifests, including explainers and podcast repurposing. It also claims stage checkpoints, render self-review, and cost previews before paid calls. The project is presented as open source under AGPL.