Cinematic narrated videos, from a single file
Give it your images and a few lines of script. It narrates each one, times every image to the words, and adds motion, music and subtitles — then renders a finished film on your Mac.
macOS · runs locally · free & open
This whole video was rendered by Narrated Slideshow itself, from one YAML file.
What it does
- Per-scene narration timed exactly to each image — measured from the voice itself, so nothing drifts.
- Ken Burns motion, cinematic transitions, colour grading, music and subtitles, all from one YAML file.
- A local voice runs free and offline on Apple Silicon; Hebrew, Arabic and 300+ languages come from the cloud, with right-to-left subtitles done properly.
- No photos? Optionally let AI generate the scene images from your script — plug in an image API key (Gemini, OpenAI, fal.ai, Replicate) or run FLUX locally on your Mac.
Install
It installs as a Claude Code skill. Paste this into your Terminal:
curl -fsSL https://pub-3a14972684b342bb923b13a28c147669.r2.dev/install.sh | bashThen, in Claude Code:
- 1
Point it at a folder of images and describe the story, e.g. “make a narrated slideshow from the photos in ~/Pictures/trip”.
- 2
It writes the scene script, narrates it, and renders your video — with subtitles and a poster frame.
macOS only; the local English voice needs Apple Silicon. Python dependencies run in ephemeral environments — the only system requirements are ffmpeg and uv, which the installer sets up. It runs in Claude Code on your own Mac — not the chat “Skills” upload.
Using the beta? Tell us what breaks