──── by Magyk ────────────────────────────────────────────────────────────────

Make video in your
terminal.

Chat with Vincent, an AI agent that turns a product photo or plain idea into a finished video — storyboard first, render when you say go.

One Gemini API key gets the starter path running: image, video, and voice.

── TWO·MODES··ONE·TERMINAL ───────────────────────────────────────

PRODUCT·MODE

Turn one product photo into a polished ad video.

[ photo.jpg ] ──▶ edit shot ──▶ animate ──▶ [ ad.mp4 ]

STORY·MODE

Turn an idea into a reviewed, multi-scene story video.

idea ──▶ storyboard ──▶ approve ──▶ render ──▶ [ story.mp4 ]
~/projects/lighthouse $ — vincent
$
● REC   1080×1920 · 30fps lighthouse_v1.mp4 4 scenes · narrated

── AFTER·THE·RENDER ──────────────────────────────────────────────────

Keep going by conversation. Say it, and Vincent re-composes only what changed.

change music rewrite scene 2 regenerate image restyle subtitles add logo animate stills

Every edit is a new version. Nothing else is touched.

── CONSISTENCY ───────────────────────────────────────────────────────────────

Drop in a reference image. Vincent can anchor every scene to it, so the same character, outfit, or setting carries through the whole story.

── YOUR·MODELS ──────────────────────────────────────────────────────────

Start simple with Gemini. Swap models when you care. Google Gemini + Veo are the zero-config default; advanced users can opt into Replicate, fal, ElevenLabs, or OpenAI — per operation.

Your keys stay on your computer — your budgets, pay as you go, straight to your inference providers.

── SCRIPT·IT ───────────────────────────────────────────────────────────

Prefer the command line, or wiring video into CI? vincent make renders without the chat — and saves a resumable session you can reopen and keep editing.

$ vincent make story "a 20s story about a lighthouse keeper"
$ vincent make storyboard ./storyboard.json
$ vincent make product ./photo.jpg "on a model in a sunlit loft"