cinetty

cinetty is an open-source terminal animation tool created by Joseph Zhou. It provides pre-choreographed animation templates that render architecture diagrams directly in terminal cells, driven by one YAML project file. Photo templates use the Kitty graphics protocol for image assets.

What It Is

cinetty works like a curated pack of motion templates for the terminal. Each template has a fixed coordinate grid, color palette, timing, and named content slots. Users supply components, values, and timed events in a YAML file, while the layout stays locked.

Agent-first workflow

The README describes a workflow where an AI coding agent inspects a codebase, picks a template, writes the project YAML, runs make check until layout constraints pass, and then plays the result. The repository root includes a SKILL.md so the clone can operate as a skill for Claude Code and OpenAI Codex, and AGENTS.md documents the equivalent workflow for other agents.

Templates and constraint checking

Six templates are listed: loop-landscape, loop-portrait, photo-loop, pipeline, screen-flow, and circuit-flow. They illustrate concepts such as self-improving agent loops, agent pipelines with routers and workers, phone-operating agents, and module circuit flows. The checker measures text in terminal cells (CJK characters count as two) and reports the exact field path and overflow, without auto-truncating text.

Setup path

It runs on macOS or Linux with Bash and Make, and requires Python 3.12+ or uv. make setup installs text-rendering dependencies (about 73 MB), while make setup-photo adds PyTorch and vision models (about 2 GB) for the photo template, which needs a Kitty-protocol terminal such as Ghostty, Kitty, or WezTerm. Playback uses make play INPUT=..., with optional looping.