tandryukha/aidemo Demo Video

Re-render a narrated, captioned product-demo MP4 in CI from a committed storyboard. Deterministic replay, no API key, ~$0.

Trust Signals

name	description	required	default
demos	Demo directories to (re-)render, space- or newline-separated. Shell globs are expanded (e.g. "demos/*"). Each dir must contain generated/storyboard.json.	yes	—
command	aidemo pipeline command: render (default, full pipeline) \| probe (selector dry-run) \| record \| compose \| voice \| captions \| gif.	no	render
tts	Voice backend. "local" runs Kokoro-82M in-process (no key, captions fall back to offline script timing); "openai" needs openai-api-key.	no	local
gif	Also export output/final-demo.gif (applies to render and compose).	no	false
headless	Run Chrome headless (required on CI runners, which have no display).	no	true
dtype	Kokoro quantization for local TTS: q8 (default, ~90 MB) \| fp32 \| fp16 \| q4 \| q4f16. Passed through as AIDEMO_TTS_MODEL_DTYPE.	no	q8
engine-ref	Engine git ref to install from github:tandryukha/aidemo (tag, branch or SHA). Defaults to the moving "stable" release pointer.	no	stable
engine-path	Use an already-checked-out engine at this path instead of installing from GitHub. Its dependencies must already be installed (e.g. `npm ci`), and for tts=local kokoro-js must be resolvable from it. Mainly for the engine's own self-test; leave empty for consumer use.	no	""
openai-api-key	OpenAI API key — only used when tts=openai. Pass it from a repository secret via the caller's `with:` block; never hard-code it.	no	""
working-directory	Directory to resolve the demo paths from.	no	.

name	description
media	Newline-separated list of rendered media files (mp4/gif).
engine-bin	Absolute path to the aidemo CLI entry the action ran.