rion-wu-tech/grok-video-workflow

Grok Video Workflow is a small command-line tool that wraps the xAI Grok Imagine Video API into a more practical local workflow. The xAI API is xAI's paid developer interface for generating short videos from text or from a reference image, and the README is clear that this project uses that billed API, not the free Grok web app quota. To use it you need xAI API credits with billing enabled in the xAI Console. The core workflow is: write a prompt, run a CLI command, and the tool calls xAI to start a generation job, polls until the job finishes, downloads the resulting MP4 to an outputs folder, and writes a JSON metadata file alongside it. A second command generates a contact sheet of still frames from the video so you can scan it for the usual generative-video problems like inconsistent identity, broken hands, jumpy scenes, or unstable on-screen text. Generation cost estimates are printed up front, and the README quotes xAI's pricing at the time of writing as roughly $0.05 per second at 480p and $0.07 per second at 720p, with a reminder to check the official console before any large batch. Text-to-video is the basic mode, and reference-to-video is the other: you can pass one or more reference images, either as local files or HTTPS URLs, and the model uses them as a visual anchor. Reference-to-video clips are capped at 10 seconds, prompts are limited to 4096 characters, and resolution is either 480p or 720p depending on what you pass, with several aspect ratios from 16:9 down to 3:2 and 2:3. The project is designed to work alongside Codex, OpenAI's terminal coding agent. The README pitches a workflow where you ask Codex to compress your prompt under the character limit, run the CLI from your terminal, look at the contact sheet, and decide whether to rerun. There is no Codex plugin involved: this is just a local tool that Codex can operate the way it would operate any other local script. Setup is the standard Node project flow with git clone, npm install, and copying a .env.example file to .env and filling in XAI_API_KEY. The README closes with a safety note asking users to respect likeness, trademark, and event broadcast rights, and not to pass off AI-generated footage as real.